---
_id: '14364'
abstract:
- lang: eng
text: We introduce extension-based proofs, a class of impossibility proofs that
includes valency arguments. They are modelled as an interaction between a prover
and a protocol. Using proofs based on combinatorial topology, it has been shown
that it is impossible to deterministically solve -set agreement among processes
or approximate agreement on a cycle of length 4 among processes in a wait-free
manner in asynchronous models where processes communicate using objects that can
be constructed from shared registers. However, it was unknown whether proofs based
on simpler techniques were possible. We show that these impossibility results
cannot be obtained by extension-based proofs in the iterated snapshot model and,
hence, extension-based proofs are limited in power.
acknowledgement: "We would like to thank Valerie King, Toniann Pitassi, and Michael
Saks for helpful discussions and Shi Hao Liu for his useful feedback.\r\nThis research
was supported by the Natural Science and Engineering Research Council of Canada
under grants RGPIN-2015-05080 and RGPIN-2020-04178, a postgraduate scholarship,
and a postdoctoral fellowship; a University of Toronto postdoctoral fellowship;
the National Science Foundation under grants CCF-1217921, CCF-1301926, CCF-1637385,
CCF-1650596, and IIS-1447786; the U.S. Department of Energy under grant ER26116/DE-SC0008923;
the European Research Council (ERC) under the European Union’s Horizon 2020 research
and innovation programme grant agreement 805223 ScaleML; and the Oracle and Intel
corporations. Some of the work on this paper was done while Faith Ellen was visiting
IST Austria."
article_processing_charge: No
article_type: original
arxiv: 1
author:
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
- first_name: James
full_name: Aspnes, James
last_name: Aspnes
- first_name: Faith
full_name: Ellen, Faith
last_name: Ellen
- first_name: Rati
full_name: Gelashvili, Rati
last_name: Gelashvili
- first_name: Leqi
full_name: Zhu, Leqi
id: a2117c59-cee4-11ed-b9d0-874ecf0f8ac5
last_name: Zhu
citation:
ama: Alistarh D-A, Aspnes J, Ellen F, Gelashvili R, Zhu L. Why extension-based proofs
fail. SIAM Journal on Computing. 2023;52(4):913-944. doi:10.1137/20M1375851
apa: Alistarh, D.-A., Aspnes, J., Ellen, F., Gelashvili, R., & Zhu, L. (2023).
Why extension-based proofs fail. SIAM Journal on Computing. Society for
Industrial and Applied Mathematics. https://doi.org/10.1137/20M1375851
chicago: Alistarh, Dan-Adrian, James Aspnes, Faith Ellen, Rati Gelashvili, and Leqi
Zhu. “Why Extension-Based Proofs Fail.” SIAM Journal on Computing. Society
for Industrial and Applied Mathematics, 2023. https://doi.org/10.1137/20M1375851.
ieee: D.-A. Alistarh, J. Aspnes, F. Ellen, R. Gelashvili, and L. Zhu, “Why extension-based
proofs fail,” SIAM Journal on Computing, vol. 52, no. 4. Society for Industrial
and Applied Mathematics, pp. 913–944, 2023.
ista: Alistarh D-A, Aspnes J, Ellen F, Gelashvili R, Zhu L. 2023. Why extension-based
proofs fail. SIAM Journal on Computing. 52(4), 913–944.
mla: Alistarh, Dan-Adrian, et al. “Why Extension-Based Proofs Fail.” SIAM Journal
on Computing, vol. 52, no. 4, Society for Industrial and Applied Mathematics,
2023, pp. 913–44, doi:10.1137/20M1375851.
short: D.-A. Alistarh, J. Aspnes, F. Ellen, R. Gelashvili, L. Zhu, SIAM Journal
on Computing 52 (2023) 913–944.
date_created: 2023-09-24T22:01:11Z
date_published: 2023-07-25T00:00:00Z
date_updated: 2023-12-13T12:28:29Z
day: '25'
department:
- _id: DaAl
doi: 10.1137/20M1375851
ec_funded: 1
external_id:
arxiv:
- '1811.01421'
isi:
- '001082972300004'
intvolume: ' 52'
isi: 1
issue: '4'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/1811.01421
month: '07'
oa: 1
oa_version: Preprint
page: 913-944
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: SIAM Journal on Computing
publication_identifier:
eissn:
- 1095-7111
issn:
- 0097-5397
publication_status: published
publisher: Society for Industrial and Applied Mathematics
quality_controlled: '1'
related_material:
record:
- id: '6676'
relation: earlier_version
status: public
scopus_import: '1'
status: public
title: Why extension-based proofs fail
type: journal_article
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 52
year: '2023'
...
---
_id: '14458'
abstract:
- lang: eng
text: 'We show for the first time that large-scale generative pretrained transformer
(GPT) family models can be pruned to at least 50% sparsity in one-shot, without
any retraining, at minimal loss of accuracy. This is achieved via a new pruning
method called SparseGPT, specifically designed to work efficiently and accurately
on massive GPT-family models. We can execute SparseGPT on the largest available
open-source models, OPT-175B and BLOOM-176B, in under 4.5 hours, and can reach
60% unstructured sparsity with negligible increase in perplexity: remarkably,
more than 100 billion weights from these models can be ignored at inference time.
SparseGPT generalizes to semi-structured (2:4 and 4:8) patterns, and is compatible
with weight quantization approaches. The code is available at: https://github.com/IST-DASLab/sparsegpt.'
acknowledged_ssus:
- _id: ScienComp
acknowledgement: The authors gratefully acknowledge funding from the European Research
Council (ERC) under the European Union’s Horizon 2020 programme (grant agreement
No. 805223 ScaleML), as well as experimental support from Eldar Kurtic, and from
the IST Austria IT department, in particular Stefano Elefante, Andrei Hornoiu, and
Alois Schloegl.
alternative_title:
- PMLR
article_processing_charge: No
arxiv: 1
author:
- first_name: Elias
full_name: Frantar, Elias
id: 09a8f98d-ec99-11ea-ae11-c063a7b7fe5f
last_name: Frantar
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Frantar E, Alistarh D-A. SparseGPT: Massive language models can be accurately
pruned in one-shot. In: Proceedings of the 40th International Conference on
Machine Learning. Vol 202. ML Research Press; 2023:10323-10337.'
apa: 'Frantar, E., & Alistarh, D.-A. (2023). SparseGPT: Massive language models
can be accurately pruned in one-shot. In Proceedings of the 40th International
Conference on Machine Learning (Vol. 202, pp. 10323–10337). Honolulu, Hawaii,
HI, United States: ML Research Press.'
chicago: 'Frantar, Elias, and Dan-Adrian Alistarh. “SparseGPT: Massive Language
Models Can Be Accurately Pruned in One-Shot.” In Proceedings of the 40th International
Conference on Machine Learning, 202:10323–37. ML Research Press, 2023.'
ieee: 'E. Frantar and D.-A. Alistarh, “SparseGPT: Massive language models can be
accurately pruned in one-shot,” in Proceedings of the 40th International Conference
on Machine Learning, Honolulu, Hawaii, HI, United States, 2023, vol. 202,
pp. 10323–10337.'
ista: 'Frantar E, Alistarh D-A. 2023. SparseGPT: Massive language models can be
accurately pruned in one-shot. Proceedings of the 40th International Conference
on Machine Learning. ICML: International Conference on Machine Learning, PMLR,
vol. 202, 10323–10337.'
mla: 'Frantar, Elias, and Dan-Adrian Alistarh. “SparseGPT: Massive Language Models
Can Be Accurately Pruned in One-Shot.” Proceedings of the 40th International
Conference on Machine Learning, vol. 202, ML Research Press, 2023, pp. 10323–37.'
short: E. Frantar, D.-A. Alistarh, in:, Proceedings of the 40th International Conference
on Machine Learning, ML Research Press, 2023, pp. 10323–10337.
conference:
end_date: 2023-07-29
location: Honolulu, Hawaii, HI, United States
name: 'ICML: International Conference on Machine Learning'
start_date: 2023-07-23
date_created: 2023-10-29T23:01:16Z
date_published: 2023-07-30T00:00:00Z
date_updated: 2023-10-31T09:59:42Z
day: '30'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2301.00774'
intvolume: ' 202'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2301.00774
month: '07'
oa: 1
oa_version: Preprint
page: 10323-10337
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the 40th International Conference on Machine Learning
publication_identifier:
eissn:
- 2640-3498
publication_status: published
publisher: ML Research Press
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'SparseGPT: Massive language models can be accurately pruned in one-shot'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 202
year: '2023'
...
---
_id: '14460'
abstract:
- lang: eng
text: We provide an efficient implementation of the backpropagation algorithm, specialized
to the case where the weights of the neural network being trained are sparse.
Our algorithm is general, as it applies to arbitrary (unstructured) sparsity and
common layer types (e.g., convolutional or linear). We provide a fast vectorized
implementation on commodity CPUs, and show that it can yield speedups in end-to-end
runtime experiments, both in transfer learning using already-sparsified networks,
and in training sparse networks from scratch. Thus, our results provide the first
support for sparse training on commodity hardware.
acknowledgement: 'We would like to thank Elias Frantar for his valuable assistance
and support at the outset of this project, and the anonymous ICML and SNN reviewers
for very constructive feedback. EI was supported in part by the FWF DK VGSCO, grant
agreement number W1260-N35. DA acknowledges generous ERC support, via Starting Grant
805223 ScaleML. '
alternative_title:
- PMLR
article_processing_charge: No
arxiv: 1
author:
- first_name: Mahdi
full_name: Nikdan, Mahdi
id: 66374281-f394-11eb-9cf6-869147deecc0
last_name: Nikdan
- first_name: Tommaso
full_name: Pegolotti, Tommaso
last_name: Pegolotti
- first_name: Eugenia B
full_name: Iofinova, Eugenia B
id: f9a17499-f6e0-11ea-865d-fdf9a3f77117
last_name: Iofinova
orcid: 0000-0002-7778-3221
- first_name: Eldar
full_name: Kurtic, Eldar
id: 47beb3a5-07b5-11eb-9b87-b108ec578218
last_name: Kurtic
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Nikdan M, Pegolotti T, Iofinova EB, Kurtic E, Alistarh D-A. SparseProp: Efficient
sparse backpropagation for faster training of neural networks at the edge. In:
Proceedings of the 40th International Conference on Machine Learning. Vol
202. ML Research Press; 2023:26215-26227.'
apa: 'Nikdan, M., Pegolotti, T., Iofinova, E. B., Kurtic, E., & Alistarh, D.-A.
(2023). SparseProp: Efficient sparse backpropagation for faster training of neural
networks at the edge. In Proceedings of the 40th International Conference on
Machine Learning (Vol. 202, pp. 26215–26227). Honolulu, Hawaii, HI, United
States: ML Research Press.'
chicago: 'Nikdan, Mahdi, Tommaso Pegolotti, Eugenia B Iofinova, Eldar Kurtic, and
Dan-Adrian Alistarh. “SparseProp: Efficient Sparse Backpropagation for Faster
Training of Neural Networks at the Edge.” In Proceedings of the 40th International
Conference on Machine Learning, 202:26215–27. ML Research Press, 2023.'
ieee: 'M. Nikdan, T. Pegolotti, E. B. Iofinova, E. Kurtic, and D.-A. Alistarh, “SparseProp:
Efficient sparse backpropagation for faster training of neural networks at the
edge,” in Proceedings of the 40th International Conference on Machine Learning,
Honolulu, Hawaii, HI, United States, 2023, vol. 202, pp. 26215–26227.'
ista: 'Nikdan M, Pegolotti T, Iofinova EB, Kurtic E, Alistarh D-A. 2023. SparseProp:
Efficient sparse backpropagation for faster training of neural networks at the
edge. Proceedings of the 40th International Conference on Machine Learning. ICML:
International Conference on Machine Learning, PMLR, vol. 202, 26215–26227.'
mla: 'Nikdan, Mahdi, et al. “SparseProp: Efficient Sparse Backpropagation for Faster
Training of Neural Networks at the Edge.” Proceedings of the 40th International
Conference on Machine Learning, vol. 202, ML Research Press, 2023, pp. 26215–27.'
short: M. Nikdan, T. Pegolotti, E.B. Iofinova, E. Kurtic, D.-A. Alistarh, in:, Proceedings
of the 40th International Conference on Machine Learning, ML Research Press, 2023,
pp. 26215–26227.
conference:
end_date: 2023-07-29
location: Honolulu, Hawaii, HI, United States
name: 'ICML: International Conference on Machine Learning'
start_date: 2023-07-23
date_created: 2023-10-29T23:01:17Z
date_published: 2023-07-30T00:00:00Z
date_updated: 2023-10-31T09:33:51Z
day: '30'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2302.04852'
intvolume: ' 202'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2302.04852
month: '07'
oa: 1
oa_version: Preprint
page: 26215-26227
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the 40th International Conference on Machine Learning
publication_identifier:
eissn:
- 2640-3498
publication_status: published
publisher: ML Research Press
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'SparseProp: Efficient sparse backpropagation for faster training of neural
networks at the edge'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 202
year: '2023'
...
---
_id: '14461'
abstract:
- lang: eng
text: 'Communication-reduction techniques are a popular way to improve scalability
in data-parallel training of deep neural networks (DNNs). The recent emergence
of large language models such as GPT has created the need for new approaches to
exploit data-parallelism. Among these, fully-sharded data parallel (FSDP) training
is highly popular, yet it still encounters scalability bottlenecks. One reason
is that applying compression techniques to FSDP is challenging: as the vast majority
of the communication involves the model’s weights, direct compression alters convergence
and leads to accuracy loss. We present QSDP, a variant of FSDP which supports
both gradient and weight quantization with theoretical guarantees, is simple to
implement and has essentially no overheads. To derive QSDP we prove that a natural
modification of SGD achieves convergence even when we only maintain quantized
weights, and thus the domain over which we train consists of quantized points
and is, therefore, highly non-convex. We validate this approach by training GPT-family
models with up to 1.3 billion parameters on a multi-node cluster. Experiments
show that QSDP preserves model accuracy, while completely removing the communication
bottlenecks of FSDP, providing end-to-end speedups of up to 2.2x.'
acknowledged_ssus:
- _id: ScienComp
acknowledgement: The authors gratefully acknowledge funding from the European Research
Council (ERC) under the European Union’s Horizon 2020 research and innovation programme
(grant agreement No 805223 ScaleML), as well as experimental support from the IST
Austria IT department, in particular Stefano Elefante, Andrei Hornoiu, and Alois
Schloegl. AV acknowledges the support of the French Agence Nationale de la Recherche
(ANR), under grant ANR-21-CE48-0016 (project COMCOPT), the support of Fondation
Hadamard with a PRMO grant, and the support of CNRS with a CoopIntEER IEA grant
(project ALFRED).
alternative_title:
- PMLR
article_processing_charge: No
arxiv: 1
author:
- first_name: Ilia
full_name: Markov, Ilia
id: D0CF4148-C985-11E9-8066-0BDEE5697425
last_name: Markov
- first_name: Adrian
full_name: Vladu, Adrian
last_name: Vladu
- first_name: Qi
full_name: Guo, Qi
last_name: Guo
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Markov I, Vladu A, Guo Q, Alistarh D-A. Quantized distributed training of
large models with convergence guarantees. In: Proceedings of the 40th International
Conference on Machine Learning. Vol 202. ML Research Press; 2023:24020-24044.'
apa: 'Markov, I., Vladu, A., Guo, Q., & Alistarh, D.-A. (2023). Quantized distributed
training of large models with convergence guarantees. In Proceedings of the
40th International Conference on Machine Learning (Vol. 202, pp. 24020–24044).
Honolulu, Hawaii, HI, United States: ML Research Press.'
chicago: Markov, Ilia, Adrian Vladu, Qi Guo, and Dan-Adrian Alistarh. “Quantized
Distributed Training of Large Models with Convergence Guarantees.” In Proceedings
of the 40th International Conference on Machine Learning, 202:24020–44. ML
Research Press, 2023.
ieee: I. Markov, A. Vladu, Q. Guo, and D.-A. Alistarh, “Quantized distributed training
of large models with convergence guarantees,” in Proceedings of the 40th International
Conference on Machine Learning, Honolulu, Hawaii, HI, United States, 2023,
vol. 202, pp. 24020–24044.
ista: 'Markov I, Vladu A, Guo Q, Alistarh D-A. 2023. Quantized distributed training
of large models with convergence guarantees. Proceedings of the 40th International
Conference on Machine Learning. ICML: International Conference on Machine Learning,
PMLR, vol. 202, 24020–24044.'
mla: Markov, Ilia, et al. “Quantized Distributed Training of Large Models with Convergence
Guarantees.” Proceedings of the 40th International Conference on Machine Learning,
vol. 202, ML Research Press, 2023, pp. 24020–44.
short: I. Markov, A. Vladu, Q. Guo, D.-A. Alistarh, in:, Proceedings of the 40th
International Conference on Machine Learning, ML Research Press, 2023, pp. 24020–24044.
conference:
end_date: 2023-07-29
location: Honolulu, Hawaii, HI, United States
name: 'ICML: International Conference on Machine Learning'
start_date: 2023-07-23
date_created: 2023-10-29T23:01:17Z
date_published: 2023-07-30T00:00:00Z
date_updated: 2023-10-31T09:40:45Z
day: '30'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2302.02390'
intvolume: ' 202'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2302.02390
month: '07'
oa: 1
oa_version: Preprint
page: 24020-24044
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the 40th International Conference on Machine Learning
publication_identifier:
eissn:
- 2640-3498
publication_status: published
publisher: ML Research Press
quality_controlled: '1'
scopus_import: '1'
status: public
title: Quantized distributed training of large models with convergence guarantees
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 202
year: '2023'
...
---
_id: '14771'
abstract:
- lang: eng
text: Pruning—that is, setting a significant subset of the parameters of a neural
network to zero—is one of the most popular methods of model compression. Yet,
several recent works have raised the issue that pruning may induce or exacerbate
bias in the output of the compressed model. Despite existing evidence for this
phenomenon, the relationship between neural network pruning and induced bias is
not well-understood. In this work, we systematically investigate and characterize
this phenomenon in Convolutional Neural Networks for computer vision. First, we
show that it is in fact possible to obtain highly-sparse models, e.g. with less
than 10% remaining weights, which do not decrease in accuracy nor substantially
increase in bias when compared to dense models. At the same time, we also find
that, at higher sparsities, pruned models exhibit higher uncertainty in their
outputs, as well as increased correlations, which we directly link to increased
bias. We propose easy-to-use criteria which, based only on the uncompressed model,
establish whether bias will increase with pruning, and identify the samples most
susceptible to biased predictions post-compression. Our code can be found at https://github.com/IST-DASLab/pruned-vision-model-bias.
acknowledgement: The authors would like to sincerely thank Sara Hooker for her feedback
during the development of this work. EI was supported in part by the FWF DK VGSCO,
grant agreement number W1260-N35. AP and DA acknowledge generous ERC support, via
Starting Grant 805223 ScaleML.
article_processing_charge: No
arxiv: 1
author:
- first_name: Eugenia B
full_name: Iofinova, Eugenia B
id: f9a17499-f6e0-11ea-865d-fdf9a3f77117
last_name: Iofinova
orcid: 0000-0002-7778-3221
- first_name: Elena-Alexandra
full_name: Peste, Elena-Alexandra
id: 32D78294-F248-11E8-B48F-1D18A9856A87
last_name: Peste
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Iofinova EB, Peste E-A, Alistarh D-A. Bias in pruned vision models: In-depth
analysis and countermeasures. In: 2023 IEEE/CVF Conference on Computer Vision
and Pattern Recognition. IEEE; 2023:24364-24373. doi:10.1109/cvpr52729.2023.02334'
apa: 'Iofinova, E. B., Peste, E.-A., & Alistarh, D.-A. (2023). Bias in pruned
vision models: In-depth analysis and countermeasures. In 2023 IEEE/CVF Conference
on Computer Vision and Pattern Recognition (pp. 24364–24373). Vancouver, BC,
Canada: IEEE. https://doi.org/10.1109/cvpr52729.2023.02334'
chicago: 'Iofinova, Eugenia B, Elena-Alexandra Peste, and Dan-Adrian Alistarh. “Bias
in Pruned Vision Models: In-Depth Analysis and Countermeasures.” In 2023 IEEE/CVF
Conference on Computer Vision and Pattern Recognition, 24364–73. IEEE, 2023.
https://doi.org/10.1109/cvpr52729.2023.02334.'
ieee: 'E. B. Iofinova, E.-A. Peste, and D.-A. Alistarh, “Bias in pruned vision models:
In-depth analysis and countermeasures,” in 2023 IEEE/CVF Conference on Computer
Vision and Pattern Recognition, Vancouver, BC, Canada, 2023, pp. 24364–24373.'
ista: 'Iofinova EB, Peste E-A, Alistarh D-A. 2023. Bias in pruned vision models:
In-depth analysis and countermeasures. 2023 IEEE/CVF Conference on Computer Vision
and Pattern Recognition. CVPR: Conference on Computer Vision and Pattern Recognition,
24364–24373.'
mla: 'Iofinova, Eugenia B., et al. “Bias in Pruned Vision Models: In-Depth Analysis
and Countermeasures.” 2023 IEEE/CVF Conference on Computer Vision and Pattern
Recognition, IEEE, 2023, pp. 24364–73, doi:10.1109/cvpr52729.2023.02334.'
short: E.B. Iofinova, E.-A. Peste, D.-A. Alistarh, in:, 2023 IEEE/CVF Conference
on Computer Vision and Pattern Recognition, IEEE, 2023, pp. 24364–24373.
conference:
end_date: 2023-06-24
location: Vancouver, BC, Canada
name: 'CVPR: Conference on Computer Vision and Pattern Recognition'
start_date: 2023-06-17
date_created: 2024-01-10T08:42:40Z
date_published: 2023-08-22T00:00:00Z
date_updated: 2024-01-10T08:59:26Z
day: '22'
department:
- _id: DaAl
- _id: ChLa
doi: 10.1109/cvpr52729.2023.02334
ec_funded: 1
external_id:
arxiv:
- '2304.12622'
isi:
- '001062531308068'
isi: 1
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2304.12622
month: '08'
oa: 1
oa_version: Preprint
page: 24364-24373
project:
- _id: 9B9290DE-BA93-11EA-9121-9846C619BF3A
grant_number: ' W1260-N35'
name: Vienna Graduate School on Computational Optimization
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition
publication_identifier:
eisbn:
- '9798350301298'
eissn:
- 2575-7075
publication_status: published
publisher: IEEE
quality_controlled: '1'
related_material:
link:
- relation: software
url: https://github.com/IST-DASLab/pruned-vision-model-bias
status: public
title: 'Bias in pruned vision models: In-depth analysis and countermeasures'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '13053'
abstract:
- lang: eng
text: 'Deep neural networks (DNNs) often have to be compressed, via pruning and/or
quantization, before they can be deployed in practical settings. In this work
we propose a new compression-aware minimizer dubbed CrAM that modifies the optimization
step in a principled way, in order to produce models whose local loss behavior
is stable under compression operations such as pruning. Thus, dense models trained
via CrAM should be compressible post-training, in a single step, without significant
accuracy loss. Experimental results on standard benchmarks, such as residual networks
for ImageNet classification and BERT models for language modelling, show that
CrAM produces dense models that can be more accurate than the standard SGD/Adam-based
baselines, but which are stable under weight pruning: specifically, we can prune
models in one-shot to 70-80% sparsity with almost no accuracy loss, and to 90%
with reasonable (∼1%) accuracy loss, which is competitive with gradual compression
methods. Additionally, CrAM can produce sparse models which perform well for transfer
learning, and it also works for semi-structured 2:4 pruning patterns supported
by GPU hardware. The code for reproducing the results is available at this https
URL .'
acknowledged_ssus:
- _id: ScienComp
acknowledgement: "AP, EK, DA received funding from the European Research Council (ERC)
under the European\r\nUnion’s Horizon 2020 research and innovation programme (grant
agreement No 805223 ScaleML). AV acknowledges the support of the French Agence Nationale
de la Recherche (ANR), under grant ANR-21-CE48-0016 (project COMCOPT). We further
acknowledge the support from the Scientific Service Units (SSU) of ISTA through
resources provided by Scientific Computing (SciComp)-"
article_processing_charge: No
arxiv: 1
author:
- first_name: Elena-Alexandra
full_name: Peste, Elena-Alexandra
id: 32D78294-F248-11E8-B48F-1D18A9856A87
last_name: Peste
- first_name: Adrian
full_name: Vladu, Adrian
last_name: Vladu
- first_name: Eldar
full_name: Kurtic, Eldar
id: 47beb3a5-07b5-11eb-9b87-b108ec578218
last_name: Kurtic
- first_name: Christoph
full_name: Lampert, Christoph
id: 40C20FD2-F248-11E8-B48F-1D18A9856A87
last_name: Lampert
orcid: 0000-0001-8622-7887
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Peste E-A, Vladu A, Kurtic E, Lampert C, Alistarh D-A. CrAM: A Compression-Aware
Minimizer. In: 11th International Conference on Learning Representations .'
apa: 'Peste, E.-A., Vladu, A., Kurtic, E., Lampert, C., & Alistarh, D.-A. (n.d.).
CrAM: A Compression-Aware Minimizer. In 11th International Conference on Learning
Representations . Kigali, Rwanda .'
chicago: 'Peste, Elena-Alexandra, Adrian Vladu, Eldar Kurtic, Christoph Lampert,
and Dan-Adrian Alistarh. “CrAM: A Compression-Aware Minimizer.” In 11th International
Conference on Learning Representations , n.d.'
ieee: 'E.-A. Peste, A. Vladu, E. Kurtic, C. Lampert, and D.-A. Alistarh, “CrAM:
A Compression-Aware Minimizer,” in 11th International Conference on Learning
Representations , Kigali, Rwanda .'
ista: 'Peste E-A, Vladu A, Kurtic E, Lampert C, Alistarh D-A. CrAM: A Compression-Aware
Minimizer. 11th International Conference on Learning Representations . ICLR: International
Conference on Learning Representations.'
mla: 'Peste, Elena-Alexandra, et al. “CrAM: A Compression-Aware Minimizer.” 11th
International Conference on Learning Representations .'
short: E.-A. Peste, A. Vladu, E. Kurtic, C. Lampert, D.-A. Alistarh, in:, 11th International
Conference on Learning Representations , n.d.
conference:
end_date: 2023-05-05
location: 'Kigali, Rwanda '
name: 'ICLR: International Conference on Learning Representations'
start_date: 2023-05-01
date_created: 2023-05-23T11:36:18Z
date_published: 2023-05-01T00:00:00Z
date_updated: 2023-06-01T12:54:45Z
department:
- _id: GradSch
- _id: DaAl
- _id: ChLa
ec_funded: 1
external_id:
arxiv:
- '2207.14200'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://openreview.net/pdf?id=_eTZBs-yedr
month: '05'
oa: 1
oa_version: Preprint
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: '11th International Conference on Learning Representations '
publication_status: accepted
quality_controlled: '1'
related_material:
record:
- id: '13074'
relation: dissertation_contains
status: public
status: public
title: 'CrAM: A Compression-Aware Minimizer'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '13074'
abstract:
- lang: eng
text: "Deep learning has become an integral part of a large number of important
applications, and many of the recent breakthroughs have been enabled by the ability
to train very large models, capable to capture complex patterns and relationships
from the data. At the same time, the massive sizes of modern deep learning models
have made their deployment to smaller devices more challenging; this is particularly
important, as in many applications the users rely on accurate deep learning predictions,
but they only have access to devices with limited memory and compute power. One
solution to this problem is to prune neural networks, by setting as many of their
parameters as possible to zero, to obtain accurate sparse models with lower memory
footprint. Despite the great research progress in obtaining sparse models that
preserve accuracy, while satisfying memory and computational constraints, there
are still many challenges associated with efficiently training sparse models,
as well as understanding their generalization properties.\r\n\r\nThe focus of
this thesis is to investigate how the training process of sparse models can be
made more efficient, and to understand the differences between sparse and dense
models in terms of how well they can generalize to changes in the data distribution.
We first study a method for co-training sparse and dense models, at a lower cost
compared to regular training. With our method we can obtain very accurate sparse
networks, and dense models that can recover the baseline accuracy. Furthermore,
we are able to more easily analyze the differences, at prediction level, between
the sparse-dense model pairs. Next, we investigate the generalization properties
of sparse neural networks in more detail, by studying how well different sparse
models trained on a larger task can adapt to smaller, more specialized tasks,
in a transfer learning scenario. Our analysis across multiple pruning methods
and sparsity levels reveals that sparse models provide features that can transfer
similarly to or better than the dense baseline. However, the choice of the pruning
method plays an important role, and can influence the results when the features
are fixed (linear finetuning), or when they are allowed to adapt to the new task
(full finetuning). Using sparse models with fixed masks for finetuning on new
tasks has an important practical advantage, as it enables training neural networks
on smaller devices. However, one drawback of current pruning methods is that the
entire training cycle has to be repeated to obtain the initial sparse model, for
every sparsity target; in consequence, the entire training process is costly and
also multiple models need to be stored. In the last part of the thesis we propose
a method that can train accurate dense models that are compressible in a single
step, to multiple sparsity levels, without additional finetuning. Our method results
in sparse models that can be competitive with existing pruning methods, and which
can also successfully generalize to new tasks."
acknowledged_ssus:
- _id: ScienComp
alternative_title:
- ISTA Thesis
article_processing_charge: No
author:
- first_name: Elena-Alexandra
full_name: Peste, Elena-Alexandra
id: 32D78294-F248-11E8-B48F-1D18A9856A87
last_name: Peste
citation:
ama: Peste E-A. Efficiency and generalization of sparse neural networks. 2023. doi:10.15479/at:ista:13074
apa: Peste, E.-A. (2023). Efficiency and generalization of sparse neural networks.
Institute of Science and Technology Austria. https://doi.org/10.15479/at:ista:13074
chicago: Peste, Elena-Alexandra. “Efficiency and Generalization of Sparse Neural
Networks.” Institute of Science and Technology Austria, 2023. https://doi.org/10.15479/at:ista:13074.
ieee: E.-A. Peste, “Efficiency and generalization of sparse neural networks,” Institute
of Science and Technology Austria, 2023.
ista: Peste E-A. 2023. Efficiency and generalization of sparse neural networks.
Institute of Science and Technology Austria.
mla: Peste, Elena-Alexandra. Efficiency and Generalization of Sparse Neural Networks.
Institute of Science and Technology Austria, 2023, doi:10.15479/at:ista:13074.
short: E.-A. Peste, Efficiency and Generalization of Sparse Neural Networks, Institute
of Science and Technology Austria, 2023.
date_created: 2023-05-23T17:07:53Z
date_published: 2023-05-23T00:00:00Z
date_updated: 2023-08-04T10:33:27Z
day: '23'
ddc:
- '000'
degree_awarded: PhD
department:
- _id: GradSch
- _id: DaAl
- _id: ChLa
doi: 10.15479/at:ista:13074
ec_funded: 1
file:
- access_level: open_access
checksum: 6b3354968403cb9d48cc5a83611fb571
content_type: application/pdf
creator: epeste
date_created: 2023-05-24T16:11:16Z
date_updated: 2023-05-24T16:11:16Z
file_id: '13087'
file_name: PhD_Thesis_Alexandra_Peste_final.pdf
file_size: 2152072
relation: main_file
success: 1
- access_level: closed
checksum: 8d0df94bbcf4db72c991f22503b3fd60
content_type: application/zip
creator: epeste
date_created: 2023-05-24T16:12:59Z
date_updated: 2023-05-24T16:12:59Z
file_id: '13088'
file_name: PhD_Thesis_APeste.zip
file_size: 1658293
relation: source_file
file_date_updated: 2023-05-24T16:12:59Z
has_accepted_license: '1'
language:
- iso: eng
month: '05'
oa: 1
oa_version: Published Version
page: '147'
project:
- _id: 2564DBCA-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '665385'
name: International IST Doctoral Program
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication_identifier:
issn:
- 2663-337X
publication_status: published
publisher: Institute of Science and Technology Austria
related_material:
record:
- id: '11458'
relation: part_of_dissertation
status: public
- id: '13053'
relation: part_of_dissertation
status: public
- id: '12299'
relation: part_of_dissertation
status: public
status: public
supervisor:
- first_name: Christoph
full_name: Lampert, Christoph
id: 40C20FD2-F248-11E8-B48F-1D18A9856A87
last_name: Lampert
orcid: 0000-0001-8622-7887
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
title: Efficiency and generalization of sparse neural networks
type: dissertation
user_id: 8b945eb4-e2f2-11eb-945a-df72226e66a9
year: '2023'
...
---
_id: '12566'
abstract:
- lang: eng
text: "Approximate agreement is one of the few variants of consensus that can be
solved in a wait-free manner in asynchronous systems where processes communicate
by reading and writing to shared memory. In this work, we consider a natural generalisation
of approximate agreement on arbitrary undirected connected graphs. Each process
is given a node of the graph as input and, if non-faulty, must output a node such
that\r\n– all the outputs are within distance 1 of one another, and\r\n– each
output value lies on a shortest path between two input values.\r\nFrom prior work,
it is known that there is no wait-free algorithm among processes for this problem
on any cycle of length , by reduction from 2-set agreement (Castañeda et al.,
2018).\r\n\r\nIn this work, we investigate the solvability of this task on general
graphs. We give a new, direct proof of the impossibility of approximate agreement
on cycles of length , via a generalisation of Sperner's Lemma to convex polygons.
We also extend the reduction from 2-set agreement to a larger class of graphs,
showing that approximate agreement on these graphs is unsolvable. On the positive
side, we present a wait-free algorithm for a different class of graphs, which
properly contains the class of chordal graphs."
acknowledgement: This project has received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 research and innovation programme
(grant agreement No. 805223 ScaleML) and under the Marie Skłodowska-Curie grant
agreement No. 840605 and from the Natural Sciences and Engineering Research Council
of Canada grant RGPIN-2020-04178. Part of this work was done while Faith Ellen was
visiting IST Austria.
article_number: '113733'
article_processing_charge: Yes (via OA deal)
article_type: original
author:
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
- first_name: Faith
full_name: Ellen, Faith
last_name: Ellen
- first_name: Joel
full_name: Rybicki, Joel
id: 334EFD2E-F248-11E8-B48F-1D18A9856A87
last_name: Rybicki
orcid: 0000-0002-6432-6646
citation:
ama: Alistarh D-A, Ellen F, Rybicki J. Wait-free approximate agreement on graphs.
Theoretical Computer Science. 2023;948(2). doi:10.1016/j.tcs.2023.113733
apa: Alistarh, D.-A., Ellen, F., & Rybicki, J. (2023). Wait-free approximate
agreement on graphs. Theoretical Computer Science. Elsevier. https://doi.org/10.1016/j.tcs.2023.113733
chicago: Alistarh, Dan-Adrian, Faith Ellen, and Joel Rybicki. “Wait-Free Approximate
Agreement on Graphs.” Theoretical Computer Science. Elsevier, 2023. https://doi.org/10.1016/j.tcs.2023.113733.
ieee: D.-A. Alistarh, F. Ellen, and J. Rybicki, “Wait-free approximate agreement
on graphs,” Theoretical Computer Science, vol. 948, no. 2. Elsevier, 2023.
ista: Alistarh D-A, Ellen F, Rybicki J. 2023. Wait-free approximate agreement on
graphs. Theoretical Computer Science. 948(2), 113733.
mla: Alistarh, Dan-Adrian, et al. “Wait-Free Approximate Agreement on Graphs.” Theoretical
Computer Science, vol. 948, no. 2, 113733, Elsevier, 2023, doi:10.1016/j.tcs.2023.113733.
short: D.-A. Alistarh, F. Ellen, J. Rybicki, Theoretical Computer Science 948 (2023).
date_created: 2023-02-19T23:00:55Z
date_published: 2023-02-28T00:00:00Z
date_updated: 2023-08-01T13:17:20Z
day: '28'
ddc:
- '000'
department:
- _id: DaAl
doi: 10.1016/j.tcs.2023.113733
ec_funded: 1
external_id:
isi:
- '000934262700001'
file:
- access_level: open_access
checksum: b27c5290f2f1500c403494364ee39c9f
content_type: application/pdf
creator: dernst
date_created: 2023-02-20T07:30:20Z
date_updated: 2023-02-20T07:30:20Z
file_id: '12570'
file_name: 2023_TheoreticalCompScience_Alistarh.pdf
file_size: 602333
relation: main_file
success: 1
file_date_updated: 2023-02-20T07:30:20Z
has_accepted_license: '1'
intvolume: ' 948'
isi: 1
issue: '2'
language:
- iso: eng
month: '02'
oa: 1
oa_version: Published Version
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
- _id: 26A5D39A-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '840605'
name: Coordination in constrained and natural distributed systems
publication: Theoretical Computer Science
publication_identifier:
issn:
- 0304-3975
publication_status: published
publisher: Elsevier
quality_controlled: '1'
scopus_import: '1'
status: public
title: Wait-free approximate agreement on graphs
tmp:
image: /images/cc_by.png
legal_code_url: https://creativecommons.org/licenses/by/4.0/legalcode
name: Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)
short: CC BY (4.0)
type: journal_article
user_id: 4359f0d1-fa6c-11eb-b949-802e58b17ae8
volume: 948
year: '2023'
...
---
_id: '11180'
abstract:
- lang: eng
text: "Designing and implementing efficient parallel priority schedulers is an active
research area. An intriguing proposed design is the Multi-Queue: given n threads
and m ≥ n distinct priority queues, task insertions are performed uniformly at
random, while, to delete, a thread picks two queues uniformly at random, and removes
the observed task of higher priority. This approach scales well, and has probabilistic
rank guarantees: roughly, the rank of each task removed, relative to remaining
tasks in all other queues, is O (m) in expectation. Yet, the performance of this
pattern is below that of well-engineered schedulers, which eschew theoretical
guarantees for practical efficiency.\r\n\r\nWe investigate whether it is possible
to design and implement a Multi-Queue-based task scheduler that is both highly-efficient
and has analytical guarantees. We propose a new variant called the Stealing Multi-Queue
(SMQ), a cache-efficient variant of the Multi-Queue, which leverages both queue
affinity---each thread has a local queue, from which tasks are usually removed;
but, with some probability, threads also attempt to steal higher-priority tasks
from the other queues---and task batching, that is, the processing of several
tasks in a single insert / remove step. These ideas are well-known for task scheduling
without priorities; our theoretical contribution is showing that, despite relaxations,
this design can still provide rank guarantees, which in turn implies bounds on
total work performed. We provide a general SMQ implementation which can surpass
state-of-the-art schedulers such as OBIM and PMOD in terms of performance on popular
graph-processing benchmarks. Notably, the performance improvement comes mainly
from the superior rank guarantees provided by our scheduler, confirming that analytically-reasoned
approaches can still provide performance improvements for priority task scheduling."
acknowledgement: We would like to thank the anonymous reviewers for their useful comments.
This project has received funding from the European Research Council (ERC) under
the European Union’s Horizon 2020 research and innovation programme (grant agreement
No 805223 ScaleML).
article_processing_charge: No
arxiv: 1
author:
- first_name: Anastasiia
full_name: Postnikova, Anastasiia
last_name: Postnikova
- first_name: Nikita
full_name: Koval, Nikita
id: 2F4DB10C-F248-11E8-B48F-1D18A9856A87
last_name: Koval
- first_name: Giorgi
full_name: Nadiradze, Giorgi
id: 3279A00C-F248-11E8-B48F-1D18A9856A87
last_name: Nadiradze
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Postnikova A, Koval N, Nadiradze G, Alistarh D-A. Multi-queues can be state-of-the-art
priority schedulers. In: Proceedings of the 27th ACM SIGPLAN Symposium on Principles
and Practice of Parallel Programming. Association for Computing Machinery;
2022:353-367. doi:10.1145/3503221.3508432'
apa: 'Postnikova, A., Koval, N., Nadiradze, G., & Alistarh, D.-A. (2022). Multi-queues
can be state-of-the-art priority schedulers. In Proceedings of the 27th ACM
SIGPLAN Symposium on Principles and Practice of Parallel Programming (pp.
353–367). Seoul, Republic of Korea: Association for Computing Machinery. https://doi.org/10.1145/3503221.3508432'
chicago: Postnikova, Anastasiia, Nikita Koval, Giorgi Nadiradze, and Dan-Adrian
Alistarh. “Multi-Queues Can Be State-of-the-Art Priority Schedulers.” In Proceedings
of the 27th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming,
353–67. Association for Computing Machinery, 2022. https://doi.org/10.1145/3503221.3508432.
ieee: A. Postnikova, N. Koval, G. Nadiradze, and D.-A. Alistarh, “Multi-queues can
be state-of-the-art priority schedulers,” in Proceedings of the 27th ACM SIGPLAN
Symposium on Principles and Practice of Parallel Programming, Seoul, Republic
of Korea, 2022, pp. 353–367.
ista: 'Postnikova A, Koval N, Nadiradze G, Alistarh D-A. 2022. Multi-queues can
be state-of-the-art priority schedulers. Proceedings of the 27th ACM SIGPLAN Symposium
on Principles and Practice of Parallel Programming. PPoPP: Sympopsium on Principles
and Practice of Parallel Programming, 353–367.'
mla: Postnikova, Anastasiia, et al. “Multi-Queues Can Be State-of-the-Art Priority
Schedulers.” Proceedings of the 27th ACM SIGPLAN Symposium on Principles and
Practice of Parallel Programming, Association for Computing Machinery, 2022,
pp. 353–67, doi:10.1145/3503221.3508432.
short: A. Postnikova, N. Koval, G. Nadiradze, D.-A. Alistarh, in:, Proceedings of
the 27th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming,
Association for Computing Machinery, 2022, pp. 353–367.
conference:
end_date: 2022-04-06
location: Seoul, Republic of Korea
name: 'PPoPP: Sympopsium on Principles and Practice of Parallel Programming'
start_date: 2022-04-02
date_created: 2022-04-17T22:01:46Z
date_published: 2022-04-02T00:00:00Z
date_updated: 2023-08-03T06:48:35Z
day: '02'
department:
- _id: DaAl
doi: 10.1145/3503221.3508432
ec_funded: 1
external_id:
arxiv:
- '2109.00657'
isi:
- '000883318200025'
isi: 1
language:
- iso: eng
main_file_link:
- open_access: '1'
url: ' https://doi.org/10.48550/arXiv.2109.00657'
month: '04'
oa: 1
oa_version: Preprint
page: 353-367
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the 27th ACM SIGPLAN Symposium on Principles and Practice
of Parallel Programming
publication_identifier:
isbn:
- '9781450392044'
publication_status: published
publisher: Association for Computing Machinery
quality_controlled: '1'
related_material:
record:
- id: '13076'
relation: research_data
status: public
scopus_import: '1'
status: public
title: Multi-queues can be state-of-the-art priority schedulers
type: conference
user_id: 4359f0d1-fa6c-11eb-b949-802e58b17ae8
year: '2022'
...
---
_id: '11183'
abstract:
- lang: eng
text: "Subgraph detection has recently been one of the most studied problems in
the CONGEST model of distributed computing. In this work, we study the distributed
complexity of problems closely related to subgraph detection, mainly focusing
on induced subgraph detection. The main line of this work presents lower bounds
and parameterized algorithms w.r.t structural parameters of the input graph:\r\n-
On general graphs, we give unconditional lower bounds for induced detection of
cycles and patterns of treewidth 2 in CONGEST. Moreover, by adapting reductions
from centralized parameterized complexity, we prove lower bounds in CONGEST for
detecting patterns with a 4-clique, and for induced path detection conditional
on the hardness of triangle detection in the congested clique.\r\n- On graphs
of bounded degeneracy, we show that induced paths can be detected fast in CONGEST
using techniques from parameterized algorithms, while detecting cycles and patterns
of treewidth 2 is hard.\r\n- On graphs of bounded vertex cover number, we show
that induced subgraph detection is easy in CONGEST for any pattern graph. More
specifically, we adapt a centralized parameterized algorithm for a more general
maximum common induced subgraph detection problem to the distributed setting.
In addition to these induced subgraph detection results, we study various related
problems in the CONGEST and congested clique models, including for multicolored
versions of subgraph-detection-like problems."
acknowledgement: "Amir Nikabadi: Supported by the LABEX MILYON (ANR-10-LABX-0070)
of Université de Lyon, within the program “Investissements d’Avenir” (ANR-11-IDEX-0007)
operated by the French National Research Agency (ANR). Janne H. Korhonen: Supported
by the European Research Council (ERC) under the European Union’s Horizon 2020 research
and innovation programme (grant agreement No 805223 ScaleML).\r\nWe thank François
Le Gall and Masayuki Miyamoto for sharing their work on lower bounds for induced
subgraph detection [36]."
alternative_title:
- LIPIcs
article_number: '15'
article_processing_charge: No
author:
- first_name: Amir
full_name: Nikabadi, Amir
last_name: Nikabadi
- first_name: Janne
full_name: Korhonen, Janne
id: C5402D42-15BC-11E9-A202-CA2BE6697425
last_name: Korhonen
citation:
ama: 'Nikabadi A, Korhonen J. Beyond distributed subgraph detection: Induced subgraphs,
multicolored problems and graph parameters. In: Bramas Q, Gramoli V, Milani A,
eds. 25th International Conference on Principles of Distributed Systems.
Vol 217. Schloss Dagstuhl - Leibniz-Zentrum für Informatik; 2022. doi:10.4230/LIPIcs.OPODIS.2021.15'
apa: 'Nikabadi, A., & Korhonen, J. (2022). Beyond distributed subgraph detection:
Induced subgraphs, multicolored problems and graph parameters. In Q. Bramas, V.
Gramoli, & A. Milani (Eds.), 25th International Conference on Principles
of Distributed Systems (Vol. 217). Strasbourg, France: Schloss Dagstuhl -
Leibniz-Zentrum für Informatik. https://doi.org/10.4230/LIPIcs.OPODIS.2021.15'
chicago: 'Nikabadi, Amir, and Janne Korhonen. “Beyond Distributed Subgraph Detection:
Induced Subgraphs, Multicolored Problems and Graph Parameters.” In 25th International
Conference on Principles of Distributed Systems, edited by Quentin Bramas,
Vincent Gramoli, and Alessia Milani, Vol. 217. Schloss Dagstuhl - Leibniz-Zentrum
für Informatik, 2022. https://doi.org/10.4230/LIPIcs.OPODIS.2021.15.'
ieee: 'A. Nikabadi and J. Korhonen, “Beyond distributed subgraph detection: Induced
subgraphs, multicolored problems and graph parameters,” in 25th International
Conference on Principles of Distributed Systems, Strasbourg, France, 2022,
vol. 217.'
ista: 'Nikabadi A, Korhonen J. 2022. Beyond distributed subgraph detection: Induced
subgraphs, multicolored problems and graph parameters. 25th International Conference
on Principles of Distributed Systems. OPODIS, LIPIcs, vol. 217, 15.'
mla: 'Nikabadi, Amir, and Janne Korhonen. “Beyond Distributed Subgraph Detection:
Induced Subgraphs, Multicolored Problems and Graph Parameters.” 25th International
Conference on Principles of Distributed Systems, edited by Quentin Bramas
et al., vol. 217, 15, Schloss Dagstuhl - Leibniz-Zentrum für Informatik, 2022,
doi:10.4230/LIPIcs.OPODIS.2021.15.'
short: A. Nikabadi, J. Korhonen, in:, Q. Bramas, V. Gramoli, A. Milani (Eds.), 25th
International Conference on Principles of Distributed Systems, Schloss Dagstuhl
- Leibniz-Zentrum für Informatik, 2022.
conference:
end_date: 2021-12-15
location: Strasbourg, France
name: OPODIS
start_date: 2021-12-13
date_created: 2022-04-17T22:01:47Z
date_published: 2022-02-01T00:00:00Z
date_updated: 2022-05-02T07:56:35Z
day: '01'
ddc:
- '510'
department:
- _id: DaAl
doi: 10.4230/LIPIcs.OPODIS.2021.15
ec_funded: 1
editor:
- first_name: Quentin
full_name: Bramas, Quentin
last_name: Bramas
- first_name: Vincent
full_name: Gramoli, Vincent
last_name: Gramoli
- first_name: Alessia
full_name: Milani, Alessia
last_name: Milani
file:
- access_level: open_access
checksum: 626551c14de5d4091573200ed0535752
content_type: application/pdf
creator: dernst
date_created: 2022-05-02T07:53:00Z
date_updated: 2022-05-02T07:53:00Z
file_id: '11345'
file_name: 2022_LIPICs_Nikabadi.pdf
file_size: 790396
relation: main_file
success: 1
file_date_updated: 2022-05-02T07:53:00Z
has_accepted_license: '1'
intvolume: ' 217'
language:
- iso: eng
month: '02'
oa: 1
oa_version: Published Version
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 25th International Conference on Principles of Distributed Systems
publication_identifier:
isbn:
- '9783959772198'
issn:
- 1868-8969
publication_status: published
publisher: Schloss Dagstuhl - Leibniz-Zentrum für Informatik
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'Beyond distributed subgraph detection: Induced subgraphs, multicolored problems
and graph parameters'
tmp:
image: /images/cc_by.png
legal_code_url: https://creativecommons.org/licenses/by/4.0/legalcode
name: Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)
short: CC BY (4.0)
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 217
year: '2022'
...
---
_id: '11184'
abstract:
- lang: eng
text: "Let G be a graph on n nodes. In the stochastic population protocol model,
a collection of n indistinguishable, resource-limited nodes collectively solve
tasks via pairwise interactions. In each interaction, two randomly chosen neighbors
first read each other’s states, and then update their local states. A rich line
of research has established tight upper and lower bounds on the complexity of
fundamental tasks, such as majority and leader election, in this model, when G
is a clique. Specifically, in the clique, these tasks can be solved fast, i.e.,
in n polylog n pairwise interactions, with high probability, using at most polylog
n states per node.\r\nIn this work, we consider the more general setting where
G is an arbitrary regular graph, and present a technique for simulating protocols
designed for fully-connected networks in any connected regular graph. Our main
result is a simulation that is efficient on many interesting graph families: roughly,
the simulation overhead is polylogarithmic in the number of nodes, and quadratic
in the conductance of the graph. As a sample application, we show that, in any
regular graph with conductance φ, both leader election and exact majority can
be solved in φ^{-2} ⋅ n polylog n pairwise interactions, with high probability,
using at most φ^{-2} ⋅ polylog n states per node. This shows that there are fast
and space-efficient population protocols for leader election and exact majority
on graphs with good expansion properties. We believe our results will prove generally
useful, as they allow efficient technology transfer between the well-mixed (clique)
case, and the under-explored spatial setting."
acknowledgement: "Dan Alistarh: This project has received funding from the European
Research Council (ERC)\r\nunder the European Union’s Horizon 2020 research and innovation
programme (grant agreement No.805223 ScaleML).\r\nJoel Rybicki: This project has
received from the European Union’s Horizon 2020 research and\r\ninnovation programme
under the Marie Skłodowska-Curie grant agreement No. 840605.\r\nAcknowledgements
We grateful to Giorgi Nadiradze for pointing out a generalisation of the phase clock
construction to non-regular graphs. We also thank anonymous reviewers for their
useful comments on earlier versions of this manuscript."
alternative_title:
- LIPIcs
article_number: '14'
article_processing_charge: No
arxiv: 1
author:
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
- first_name: Rati
full_name: Gelashvili, Rati
last_name: Gelashvili
- first_name: Joel
full_name: Rybicki, Joel
id: 334EFD2E-F248-11E8-B48F-1D18A9856A87
last_name: Rybicki
orcid: 0000-0002-6432-6646
citation:
ama: 'Alistarh D-A, Gelashvili R, Rybicki J. Fast graphical population protocols.
In: Bramas Q, Gramoli V, Milani A, eds. 25th International Conference on Principles
of Distributed Systems. Vol 217. Schloss Dagstuhl - Leibniz-Zentrum für Informatik;
2022. doi:10.4230/LIPIcs.OPODIS.2021.14'
apa: 'Alistarh, D.-A., Gelashvili, R., & Rybicki, J. (2022). Fast graphical
population protocols. In Q. Bramas, V. Gramoli, & A. Milani (Eds.), 25th
International Conference on Principles of Distributed Systems (Vol. 217).
Strasbourg, France: Schloss Dagstuhl - Leibniz-Zentrum für Informatik. https://doi.org/10.4230/LIPIcs.OPODIS.2021.14'
chicago: Alistarh, Dan-Adrian, Rati Gelashvili, and Joel Rybicki. “Fast Graphical
Population Protocols.” In 25th International Conference on Principles of Distributed
Systems, edited by Quentin Bramas, Vincent Gramoli, and Alessia Milani, Vol.
217. Schloss Dagstuhl - Leibniz-Zentrum für Informatik, 2022. https://doi.org/10.4230/LIPIcs.OPODIS.2021.14.
ieee: D.-A. Alistarh, R. Gelashvili, and J. Rybicki, “Fast graphical population
protocols,” in 25th International Conference on Principles of Distributed Systems,
Strasbourg, France, 2022, vol. 217.
ista: Alistarh D-A, Gelashvili R, Rybicki J. 2022. Fast graphical population protocols.
25th International Conference on Principles of Distributed Systems. OPODIS, LIPIcs,
vol. 217, 14.
mla: Alistarh, Dan-Adrian, et al. “Fast Graphical Population Protocols.” 25th
International Conference on Principles of Distributed Systems, edited by Quentin
Bramas et al., vol. 217, 14, Schloss Dagstuhl - Leibniz-Zentrum für Informatik,
2022, doi:10.4230/LIPIcs.OPODIS.2021.14.
short: D.-A. Alistarh, R. Gelashvili, J. Rybicki, in:, Q. Bramas, V. Gramoli, A.
Milani (Eds.), 25th International Conference on Principles of Distributed Systems,
Schloss Dagstuhl - Leibniz-Zentrum für Informatik, 2022.
conference:
end_date: 2021-12-15
location: Strasbourg, France
name: OPODIS
start_date: 2021-12-13
date_created: 2022-04-17T22:01:47Z
date_published: 2022-02-01T00:00:00Z
date_updated: 2022-05-02T08:09:39Z
day: '01'
ddc:
- '510'
department:
- _id: DaAl
doi: 10.4230/LIPIcs.OPODIS.2021.14
ec_funded: 1
editor:
- first_name: Quentin
full_name: Bramas, Quentin
last_name: Bramas
- first_name: Vincent
full_name: Gramoli, Vincent
last_name: Gramoli
- first_name: Alessia
full_name: Milani, Alessia
last_name: Milani
external_id:
arxiv:
- '2102.08808'
file:
- access_level: open_access
checksum: 2c7c982174c6f98c4ca6e92539d15086
content_type: application/pdf
creator: dernst
date_created: 2022-05-02T08:06:33Z
date_updated: 2022-05-02T08:06:33Z
file_id: '11346'
file_name: 2022_LIPICs_Alistarh.pdf
file_size: 959406
relation: main_file
success: 1
file_date_updated: 2022-05-02T08:06:33Z
has_accepted_license: '1'
intvolume: ' 217'
language:
- iso: eng
month: '02'
oa: 1
oa_version: Published Version
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
- _id: 26A5D39A-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '840605'
name: Coordination in constrained and natural distributed systems
publication: 25th International Conference on Principles of Distributed Systems
publication_identifier:
isbn:
- '9783959772198'
issn:
- 1868-8969
publication_status: published
publisher: Schloss Dagstuhl - Leibniz-Zentrum für Informatik
quality_controlled: '1'
scopus_import: '1'
status: public
title: Fast graphical population protocols
tmp:
image: /images/cc_by.png
legal_code_url: https://creativecommons.org/licenses/by/4.0/legalcode
name: Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)
short: CC BY (4.0)
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 217
year: '2022'
...
---
_id: '11844'
abstract:
- lang: eng
text: "In the stochastic population protocol model, we are given a connected graph
with n nodes, and in every time step, a scheduler samples an edge of the graph
uniformly at random and the nodes connected by this edge interact. A fundamental
task in this model is stable leader election, in which all nodes start in an identical
state and the aim is to reach a configuration in which (1) exactly one node is
elected as leader and (2) this node remains as the unique leader no matter what
sequence of interactions follows. On cliques, the complexity of this problem has
recently been settled: time-optimal protocols stabilize in Θ(n log n) expected
steps using Θ(log log n) states, whereas protocols that use O(1) states require
Θ(n2) expected steps.\r\n\r\nIn this work, we investigate the complexity of stable
leader election on general graphs. We provide the first non-trivial time lower
bounds for leader election on general graphs, showing that, when moving beyond
cliques, the complexity landscape of leader election becomes very diverse: the
time required to elect a leader can range from O(1) to Θ(n3) expected steps. On
the upper bound side, we first observe that there exists a protocol that is time-optimal
on many graph families, but uses polynomially-many states. In contrast, we give
a near-time-optimal protocol that uses only O(log2n) states that is at most a
factor log n slower. Finally, we show that the constant-state protocol of Beauquier
et al. [OPODIS 2013] is at most a factor n log n slower than the fast polynomial-state
protocol. Moreover, among constant-state protocols, this protocol has near-optimal
average case complexity on dense random graphs."
acknowledgement: We thank the anonymous reviewers for their helpful comments. We gratefully
acknowledge funding from the European Research Council (ERC) under the European
Union’s Horizon 2020 research and innovation programme (grant agreement No 805223
ScaleML).
article_processing_charge: Yes (via OA deal)
arxiv: 1
author:
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
- first_name: Joel
full_name: Rybicki, Joel
id: 334EFD2E-F248-11E8-B48F-1D18A9856A87
last_name: Rybicki
orcid: 0000-0002-6432-6646
- first_name: Sasha
full_name: Voitovych, Sasha
last_name: Voitovych
citation:
ama: 'Alistarh D-A, Rybicki J, Voitovych S. Near-optimal leader election in population
protocols on graphs. In: Proceedings of the Annual ACM Symposium on Principles
of Distributed Computing. Association for Computing Machinery; 2022:246-256.
doi:10.1145/3519270.3538435'
apa: 'Alistarh, D.-A., Rybicki, J., & Voitovych, S. (2022). Near-optimal leader
election in population protocols on graphs. In Proceedings of the Annual ACM
Symposium on Principles of Distributed Computing (pp. 246–256). Salerno, Italy:
Association for Computing Machinery. https://doi.org/10.1145/3519270.3538435'
chicago: Alistarh, Dan-Adrian, Joel Rybicki, and Sasha Voitovych. “Near-Optimal
Leader Election in Population Protocols on Graphs.” In Proceedings of the Annual
ACM Symposium on Principles of Distributed Computing, 246–56. Association
for Computing Machinery, 2022. https://doi.org/10.1145/3519270.3538435.
ieee: D.-A. Alistarh, J. Rybicki, and S. Voitovych, “Near-optimal leader election
in population protocols on graphs,” in Proceedings of the Annual ACM Symposium
on Principles of Distributed Computing, Salerno, Italy, 2022, pp. 246–256.
ista: 'Alistarh D-A, Rybicki J, Voitovych S. 2022. Near-optimal leader election
in population protocols on graphs. Proceedings of the Annual ACM Symposium on
Principles of Distributed Computing. PODC: Symposium on Principles of Distributed
Computing, 246–256.'
mla: Alistarh, Dan-Adrian, et al. “Near-Optimal Leader Election in Population Protocols
on Graphs.” Proceedings of the Annual ACM Symposium on Principles of Distributed
Computing, Association for Computing Machinery, 2022, pp. 246–56, doi:10.1145/3519270.3538435.
short: D.-A. Alistarh, J. Rybicki, S. Voitovych, in:, Proceedings of the Annual
ACM Symposium on Principles of Distributed Computing, Association for Computing
Machinery, 2022, pp. 246–256.
conference:
end_date: 2022-07-29
location: Salerno, Italy
name: 'PODC: Symposium on Principles of Distributed Computing'
start_date: 2022-07-25
date_created: 2022-08-14T22:01:46Z
date_published: 2022-07-21T00:00:00Z
date_updated: 2023-06-14T12:06:01Z
day: '21'
ddc:
- '000'
department:
- _id: DaAl
doi: 10.1145/3519270.3538435
ec_funded: 1
external_id:
arxiv:
- '2205.12597'
file:
- access_level: open_access
checksum: 4c6b29172b8e355b4fbc364a2e0827b2
content_type: application/pdf
creator: cchlebak
date_created: 2022-08-16T08:05:15Z
date_updated: 2022-08-16T08:05:15Z
file_id: '11854'
file_name: 2022_PODC_Alistarh.pdf
file_size: 1593474
relation: main_file
success: 1
file_date_updated: 2022-08-16T08:05:15Z
has_accepted_license: '1'
language:
- iso: eng
month: '07'
oa: 1
oa_version: Published Version
page: 246-256
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the Annual ACM Symposium on Principles of Distributed
Computing
publication_identifier:
isbn:
- '9781450392624'
publication_status: published
publisher: Association for Computing Machinery
quality_controlled: '1'
scopus_import: '1'
status: public
title: Near-optimal leader election in population protocols on graphs
tmp:
image: /images/cc_by.png
legal_code_url: https://creativecommons.org/licenses/by/4.0/legalcode
name: Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)
short: CC BY (4.0)
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2022'
...
---
_id: '12299'
abstract:
- lang: eng
text: 'Transfer learning is a classic paradigm by which models pretrained on large
“upstream” datasets are adapted to yield good results on “downstream” specialized
datasets. Generally, more accurate models on the “upstream” dataset tend to provide
better transfer accuracy “downstream”. In this work, we perform an in-depth investigation
of this phenomenon in the context of convolutional neural networks (CNNs) trained
on the ImageNet dataset, which have been pruned-that is, compressed by sparsifiying
their connections. We consider transfer using unstructured pruned models obtained
by applying several state-of-the-art pruning methods, including magnitude-based,
second-order, regrowth, lottery-ticket, and regularization approaches, in the
context of twelve standard transfer tasks. In a nutshell, our study shows that
sparse models can match or even outperform the transfer performance of dense models,
even at high sparsities, and, while doing so, can lead to significant inference
and even training speedups. At the same time, we observe and analyze significant
differences in the behaviour of different pruning methods. The code is available
at: https://github.com/IST-DASLab/sparse-imagenet-transfer.'
acknowledgement: he authors would like to sincerely thank Christoph Lampert and Nir
Shavit for fruitful discussions during the development of this work, and Eldar Kurtic
for experimental support. EI was supported in part by the FWF DK VGSCO, grant agreement
number W1260-N35, while AP and DA acknowledge generous support by the ERC, via Starting
Grant 805223 ScaleML.
article_processing_charge: No
arxiv: 1
author:
- first_name: Eugenia B
full_name: Iofinova, Eugenia B
id: f9a17499-f6e0-11ea-865d-fdf9a3f77117
last_name: Iofinova
orcid: 0000-0002-7778-3221
- first_name: Elena-Alexandra
full_name: Peste, Elena-Alexandra
id: 32D78294-F248-11E8-B48F-1D18A9856A87
last_name: Peste
- first_name: Mark
full_name: Kurtz, Mark
last_name: Kurtz
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Iofinova EB, Peste E-A, Kurtz M, Alistarh D-A. How well do sparse ImageNet
models transfer? In: 2022 IEEE/CVF Conference on Computer Vision and Pattern
Recognition. Institute of Electrical and Electronics Engineers; 2022:12256-12266.
doi:10.1109/cvpr52688.2022.01195'
apa: 'Iofinova, E. B., Peste, E.-A., Kurtz, M., & Alistarh, D.-A. (2022). How
well do sparse ImageNet models transfer? In 2022 IEEE/CVF Conference on Computer
Vision and Pattern Recognition (pp. 12256–12266). New Orleans, LA, United
States: Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/cvpr52688.2022.01195'
chicago: Iofinova, Eugenia B, Elena-Alexandra Peste, Mark Kurtz, and Dan-Adrian
Alistarh. “How Well Do Sparse ImageNet Models Transfer?” In 2022 IEEE/CVF Conference
on Computer Vision and Pattern Recognition, 12256–66. Institute of Electrical
and Electronics Engineers, 2022. https://doi.org/10.1109/cvpr52688.2022.01195.
ieee: E. B. Iofinova, E.-A. Peste, M. Kurtz, and D.-A. Alistarh, “How well do sparse
ImageNet models transfer?,” in 2022 IEEE/CVF Conference on Computer Vision
and Pattern Recognition, New Orleans, LA, United States, 2022, pp. 12256–12266.
ista: 'Iofinova EB, Peste E-A, Kurtz M, Alistarh D-A. 2022. How well do sparse ImageNet
models transfer? 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition.
CVPR: Computer Vision and Pattern Recognition, 12256–12266.'
mla: Iofinova, Eugenia B., et al. “How Well Do Sparse ImageNet Models Transfer?”
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Institute
of Electrical and Electronics Engineers, 2022, pp. 12256–66, doi:10.1109/cvpr52688.2022.01195.
short: E.B. Iofinova, E.-A. Peste, M. Kurtz, D.-A. Alistarh, in:, 2022 IEEE/CVF
Conference on Computer Vision and Pattern Recognition, Institute of Electrical
and Electronics Engineers, 2022, pp. 12256–12266.
conference:
end_date: 2022-06-24
location: New Orleans, LA, United States
name: 'CVPR: Computer Vision and Pattern Recognition'
start_date: 2022-06-18
date_created: 2023-01-16T10:06:00Z
date_published: 2022-09-27T00:00:00Z
date_updated: 2023-08-04T10:33:28Z
day: '27'
department:
- _id: DaAl
- _id: ChLa
doi: 10.1109/cvpr52688.2022.01195
ec_funded: 1
external_id:
arxiv:
- '2111.13445'
isi:
- '000870759105034'
isi: 1
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2111.13445
month: '09'
oa: 1
oa_version: Preprint
page: 12256-12266
project:
- _id: 9B9290DE-BA93-11EA-9121-9846C619BF3A
grant_number: ' W1260-N35'
name: Vienna Graduate School on Computational Optimization
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition
publication_identifier:
eissn:
- 2575-7075
publication_status: published
publisher: Institute of Electrical and Electronics Engineers
quality_controlled: '1'
related_material:
record:
- id: '13074'
relation: dissertation_contains
status: public
scopus_import: '1'
status: public
title: How well do sparse ImageNet models transfer?
type: conference
user_id: 4359f0d1-fa6c-11eb-b949-802e58b17ae8
year: '2022'
...
---
_id: '10854'
abstract:
- lang: eng
text: "Consider a distributed task where the communication network is fixed but
the local inputs given to the nodes of the distributed system may change over
time. In this work, we explore the following question: if some of the local inputs
change, can an existing solution be updated efficiently, in a dynamic and distributed
manner?\r\nTo address this question, we define the batch dynamic CONGEST model
in which we are given a bandwidth-limited communication network and a dynamic
edge labelling defines the problem input. The task is to maintain a solution to
a graph problem on the labelled graph under batch changes. We investigate, when
a batch of alpha edge label changes arrive, - how much time as a function of alpha
we need to update an existing solution, and - how much information the nodes have
to keep in local memory between batches in order to update the solution quickly.\r\nOur
work lays the foundations for the theory of input-dynamic distributed network
algorithms. We give a general picture of the complexity landscape in this model,
design both universal algorithms and algorithms for concrete problems, and present
a general framework for lower bounds. The diverse time complexity of our model
spans from constant time, through time polynomial in alpha, and to alpha time,
which we show to be enough for any task."
acknowledgement: We thank Jukka Suomela for discussions. We also thank our shepherd
Mohammad Hajiesmaili and the reviewers for their time and suggestions on how to
improve the paper. This project has received funding from the European Research
Council (ERC) under the European Union’s Horizon 2020 research and innovation programme
(grant agreement No 805223 ScaleML), from the European Union’s Horizon 2020 research
and innovation programme under the Marie Skłodowska–Curie grant agreement No. 840605,
from the Vienna Science and Technology Fund (WWTF) project WHATIF, ICT19-045, 2020-2024,
and from the Austrian Science Fund (FWF) and netIDEE SCIENCE project P 33775-N.
article_processing_charge: No
arxiv: 1
author:
- first_name: Klaus-Tycho
full_name: Foerster, Klaus-Tycho
last_name: Foerster
- first_name: Janne
full_name: Korhonen, Janne
id: C5402D42-15BC-11E9-A202-CA2BE6697425
last_name: Korhonen
- first_name: Ami
full_name: Paz, Ami
last_name: Paz
- first_name: Joel
full_name: Rybicki, Joel
id: 334EFD2E-F248-11E8-B48F-1D18A9856A87
last_name: Rybicki
orcid: 0000-0002-6432-6646
- first_name: Stefan
full_name: Schmid, Stefan
last_name: Schmid
citation:
ama: 'Foerster K-T, Korhonen J, Paz A, Rybicki J, Schmid S. Input-dynamic distributed
algorithms for communication networks. In: Abstract Proceedings of the 2021
ACM SIGMETRICS / International Conference on Measurement and Modeling of Computer
Systems. Association for Computing Machinery; 2021:71-72. doi:10.1145/3410220.3453923'
apa: 'Foerster, K.-T., Korhonen, J., Paz, A., Rybicki, J., & Schmid, S. (2021).
Input-dynamic distributed algorithms for communication networks. In Abstract
Proceedings of the 2021 ACM SIGMETRICS / International Conference on Measurement
and Modeling of Computer Systems (pp. 71–72). Virtual, Online: Association
for Computing Machinery. https://doi.org/10.1145/3410220.3453923'
chicago: Foerster, Klaus-Tycho, Janne Korhonen, Ami Paz, Joel Rybicki, and Stefan
Schmid. “Input-Dynamic Distributed Algorithms for Communication Networks.” In
Abstract Proceedings of the 2021 ACM SIGMETRICS / International Conference
on Measurement and Modeling of Computer Systems, 71–72. Association for Computing
Machinery, 2021. https://doi.org/10.1145/3410220.3453923.
ieee: K.-T. Foerster, J. Korhonen, A. Paz, J. Rybicki, and S. Schmid, “Input-dynamic
distributed algorithms for communication networks,” in Abstract Proceedings
of the 2021 ACM SIGMETRICS / International Conference on Measurement and Modeling
of Computer Systems, Virtual, Online, 2021, pp. 71–72.
ista: 'Foerster K-T, Korhonen J, Paz A, Rybicki J, Schmid S. 2021. Input-dynamic
distributed algorithms for communication networks. Abstract Proceedings of the
2021 ACM SIGMETRICS / International Conference on Measurement and Modeling of
Computer Systems. SIGMETRICS: International Conference on Measurement and Modeling
of Computer Systems, 71–72.'
mla: Foerster, Klaus-Tycho, et al. “Input-Dynamic Distributed Algorithms for Communication
Networks.” Abstract Proceedings of the 2021 ACM SIGMETRICS / International
Conference on Measurement and Modeling of Computer Systems, Association for
Computing Machinery, 2021, pp. 71–72, doi:10.1145/3410220.3453923.
short: K.-T. Foerster, J. Korhonen, A. Paz, J. Rybicki, S. Schmid, in:, Abstract
Proceedings of the 2021 ACM SIGMETRICS / International Conference on Measurement
and Modeling of Computer Systems, Association for Computing Machinery, 2021, pp.
71–72.
conference:
end_date: 2021-06-18
location: Virtual, Online
name: 'SIGMETRICS: International Conference on Measurement and Modeling of Computer
Systems'
start_date: 2021-06-14
date_created: 2022-03-18T08:48:41Z
date_published: 2021-05-01T00:00:00Z
date_updated: 2023-09-26T10:40:55Z
day: '01'
department:
- _id: DaAl
doi: 10.1145/3410220.3453923
ec_funded: 1
external_id:
arxiv:
- '2005.07637'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2005.07637
month: '05'
oa: 1
oa_version: Preprint
page: 71-72
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
- _id: 26A5D39A-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '840605'
name: Coordination in constrained and natural distributed systems
publication: Abstract Proceedings of the 2021 ACM SIGMETRICS / International Conference
on Measurement and Modeling of Computer Systems
publication_identifier:
isbn:
- '9781450380720'
publication_status: published
publisher: Association for Computing Machinery
quality_controlled: '1'
related_material:
record:
- id: '10855'
relation: extended_version
status: public
scopus_import: '1'
status: public
title: Input-dynamic distributed algorithms for communication networks
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2021'
...
---
_id: '10855'
abstract:
- lang: eng
text: 'Consider a distributed task where the communication network is fixed but
the local inputs given to the nodes of the distributed system may change over
time. In this work, we explore the following question: if some of the local inputs
change, can an existing solution be updated efficiently, in a dynamic and distributed
manner? To address this question, we define the batch dynamic \congest model in
which we are given a bandwidth-limited communication network and a dynamic edge
labelling defines the problem input. The task is to maintain a solution to a graph
problem on the labeled graph under batch changes. We investigate, when a batch
of α edge label changes arrive, \beginitemize \item how much time as a function
of α we need to update an existing solution, and \item how much information the
nodes have to keep in local memory between batches in order to update the solution
quickly. \enditemize Our work lays the foundations for the theory of input-dynamic
distributed network algorithms. We give a general picture of the complexity landscape
in this model, design both universal algorithms and algorithms for concrete problems,
and present a general framework for lower bounds. In particular, we derive non-trivial
upper bounds for two selected, contrasting problems: maintaining a minimum spanning
tree and detecting cliques.'
acknowledgement: "We thank Jukka Suomela for discussions. We also thank our shepherd
Mohammad Hajiesmaili\r\nand the reviewers for their time and suggestions on how
to improve the paper. This project\r\nhas received funding from the European Research
Council (ERC) under the European Union’s\r\nHorizon 2020 research and innovation
programme (grant agreement No 805223 ScaleML), from the European Union’s Horizon
2020 research and innovation programme under the Marie\r\nSk lodowska–Curie grant
agreement No. 840605, from the Vienna Science and Technology Fund (WWTF) project
WHATIF, ICT19-045, 2020-2024, and from the Austrian Science Fund (FWF) and netIDEE
SCIENCE project P 33775-N."
article_processing_charge: No
article_type: original
arxiv: 1
author:
- first_name: Klaus-Tycho
full_name: Foerster, Klaus-Tycho
last_name: Foerster
- first_name: Janne
full_name: Korhonen, Janne
id: C5402D42-15BC-11E9-A202-CA2BE6697425
last_name: Korhonen
- first_name: Ami
full_name: Paz, Ami
last_name: Paz
- first_name: Joel
full_name: Rybicki, Joel
id: 334EFD2E-F248-11E8-B48F-1D18A9856A87
last_name: Rybicki
orcid: 0000-0002-6432-6646
- first_name: Stefan
full_name: Schmid, Stefan
last_name: Schmid
citation:
ama: Foerster K-T, Korhonen J, Paz A, Rybicki J, Schmid S. Input-dynamic distributed
algorithms for communication networks. Proceedings of the ACM on Measurement
and Analysis of Computing Systems. 2021;5(1):1-33. doi:10.1145/3447384
apa: Foerster, K.-T., Korhonen, J., Paz, A., Rybicki, J., & Schmid, S. (2021).
Input-dynamic distributed algorithms for communication networks. Proceedings
of the ACM on Measurement and Analysis of Computing Systems. Association for
Computing Machinery. https://doi.org/10.1145/3447384
chicago: Foerster, Klaus-Tycho, Janne Korhonen, Ami Paz, Joel Rybicki, and Stefan
Schmid. “Input-Dynamic Distributed Algorithms for Communication Networks.” Proceedings
of the ACM on Measurement and Analysis of Computing Systems. Association for
Computing Machinery, 2021. https://doi.org/10.1145/3447384.
ieee: K.-T. Foerster, J. Korhonen, A. Paz, J. Rybicki, and S. Schmid, “Input-dynamic
distributed algorithms for communication networks,” Proceedings of the ACM
on Measurement and Analysis of Computing Systems, vol. 5, no. 1. Association
for Computing Machinery, pp. 1–33, 2021.
ista: Foerster K-T, Korhonen J, Paz A, Rybicki J, Schmid S. 2021. Input-dynamic
distributed algorithms for communication networks. Proceedings of the ACM on Measurement
and Analysis of Computing Systems. 5(1), 1–33.
mla: Foerster, Klaus-Tycho, et al. “Input-Dynamic Distributed Algorithms for Communication
Networks.” Proceedings of the ACM on Measurement and Analysis of Computing
Systems, vol. 5, no. 1, Association for Computing Machinery, 2021, pp. 1–33,
doi:10.1145/3447384.
short: K.-T. Foerster, J. Korhonen, A. Paz, J. Rybicki, S. Schmid, Proceedings of
the ACM on Measurement and Analysis of Computing Systems 5 (2021) 1–33.
date_created: 2022-03-18T09:10:27Z
date_published: 2021-03-01T00:00:00Z
date_updated: 2023-09-26T10:40:55Z
day: '01'
department:
- _id: DaAl
doi: 10.1145/3447384
ec_funded: 1
external_id:
arxiv:
- '2005.07637'
intvolume: ' 5'
issue: '1'
keyword:
- Computer Networks and Communications
- Hardware and Architecture
- Safety
- Risk
- Reliability and Quality
- Computer Science (miscellaneous)
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2005.07637
month: '03'
oa: 1
oa_version: Preprint
page: 1-33
project:
- _id: 26A5D39A-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '840605'
name: Coordination in constrained and natural distributed systems
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: Proceedings of the ACM on Measurement and Analysis of Computing Systems
publication_identifier:
issn:
- 2476-1249
publication_status: published
publisher: Association for Computing Machinery
quality_controlled: '1'
related_material:
record:
- id: '10854'
relation: shorter_version
status: public
scopus_import: '1'
status: public
title: Input-dynamic distributed algorithms for communication networks
type: journal_article
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 5
year: '2021'
...
---
_id: '11436'
abstract:
- lang: eng
text: Asynchronous distributed algorithms are a popular way to reduce synchronization
costs in large-scale optimization, and in particular for neural network training.
However, for nonsmooth and nonconvex objectives, few convergence guarantees exist
beyond cases where closed-form proximal operator solutions are available. As training
most popular deep neural networks corresponds to optimizing nonsmooth and nonconvex
objectives, there is a pressing need for such convergence guarantees. In this
paper, we analyze for the first time the convergence of stochastic asynchronous
optimization for this general class of objectives. In particular, we focus on
stochastic subgradient methods allowing for block variable partitioning, where
the shared model is asynchronously updated by concurrent processes. To this end,
we use a probabilistic model which captures key features of real asynchronous
scheduling between concurrent processes. Under this model, we establish convergence
with probability one to an invariant set for stochastic subgradient methods with
momentum. From a practical perspective, one issue with the family of algorithms
that we consider is that they are not efficiently supported by machine learning
frameworks, which mostly focus on distributed data-parallel strategies. To address
this, we propose a new implementation strategy for shared-memory based training
of deep neural networks for a partitioned but shared model in single- and multi-GPU
settings. Based on this implementation, we achieve on average1.2x speed-up in
comparison to state-of-the-art training methods for popular image classification
tasks, without compromising accuracy.
acknowledgement: Vyacheslav Kungurtsev was supported by the OP VVV project CZ.02.1.01/0.0/0.0/16
019/0000765 “Research Center for Informatics. Bapi Chatterjee was supported by the
European Union’s Horizon 2020 research and innovation programme under the Marie
Sklodowska-Curie grant agreement No. 754411 (ISTPlus). Dan Alistarh has received
funding from the European Research Council (ERC) under the European Union’s Horizon
2020 research and innovation programme (grant agreement No 805223 ScaleML).
article_processing_charge: No
arxiv: 1
author:
- first_name: Vyacheslav
full_name: Kungurtsev, Vyacheslav
last_name: Kungurtsev
- first_name: Malcolm
full_name: Egan, Malcolm
last_name: Egan
- first_name: Bapi
full_name: Chatterjee, Bapi
id: 3C41A08A-F248-11E8-B48F-1D18A9856A87
last_name: Chatterjee
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Kungurtsev V, Egan M, Chatterjee B, Alistarh D-A. Asynchronous optimization
methods for efficient training of deep neural networks with guarantees. In: 35th
AAAI Conference on Artificial Intelligence, AAAI 2021. Vol 35. AAAI Press;
2021:8209-8216.'
apa: 'Kungurtsev, V., Egan, M., Chatterjee, B., & Alistarh, D.-A. (2021). Asynchronous
optimization methods for efficient training of deep neural networks with guarantees.
In 35th AAAI Conference on Artificial Intelligence, AAAI 2021 (Vol. 35,
pp. 8209–8216). Virtual, Online: AAAI Press.'
chicago: Kungurtsev, Vyacheslav, Malcolm Egan, Bapi Chatterjee, and Dan-Adrian Alistarh.
“Asynchronous Optimization Methods for Efficient Training of Deep Neural Networks
with Guarantees.” In 35th AAAI Conference on Artificial Intelligence, AAAI
2021, 35:8209–16. AAAI Press, 2021.
ieee: V. Kungurtsev, M. Egan, B. Chatterjee, and D.-A. Alistarh, “Asynchronous optimization
methods for efficient training of deep neural networks with guarantees,” in 35th
AAAI Conference on Artificial Intelligence, AAAI 2021, Virtual, Online, 2021,
vol. 35, no. 9B, pp. 8209–8216.
ista: 'Kungurtsev V, Egan M, Chatterjee B, Alistarh D-A. 2021. Asynchronous optimization
methods for efficient training of deep neural networks with guarantees. 35th AAAI
Conference on Artificial Intelligence, AAAI 2021. AAAI: Conference on Artificial
Intelligence vol. 35, 8209–8216.'
mla: Kungurtsev, Vyacheslav, et al. “Asynchronous Optimization Methods for Efficient
Training of Deep Neural Networks with Guarantees.” 35th AAAI Conference on
Artificial Intelligence, AAAI 2021, vol. 35, no. 9B, AAAI Press, 2021, pp.
8209–16.
short: V. Kungurtsev, M. Egan, B. Chatterjee, D.-A. Alistarh, in:, 35th AAAI Conference
on Artificial Intelligence, AAAI 2021, AAAI Press, 2021, pp. 8209–8216.
conference:
end_date: 2021-02-09
location: Virtual, Online
name: 'AAAI: Conference on Artificial Intelligence'
start_date: 2021-02-02
date_created: 2022-06-05T22:01:52Z
date_published: 2021-05-18T00:00:00Z
date_updated: 2022-06-07T06:53:36Z
day: '18'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '1905.11845'
intvolume: ' 35'
issue: 9B
language:
- iso: eng
main_file_link:
- open_access: '1'
url: ' https://doi.org/10.48550/arXiv.1905.11845'
month: '05'
oa: 1
oa_version: Preprint
page: 8209-8216
project:
- _id: 260C2330-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '754411'
name: ISTplus - Postdoctoral Fellowships
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 35th AAAI Conference on Artificial Intelligence, AAAI 2021
publication_identifier:
eissn:
- 2374-3468
isbn:
- '9781713835974'
issn:
- 2159-5399
publication_status: published
publisher: AAAI Press
quality_controlled: '1'
scopus_import: '1'
status: public
title: Asynchronous optimization methods for efficient training of deep neural networks
with guarantees
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 35
year: '2021'
...
---
_id: '11452'
abstract:
- lang: eng
text: We study efficient distributed algorithms for the fundamental problem of principal
component analysis and leading eigenvector computation on the sphere, when the
data are randomly distributed among a set of computational nodes. We propose a
new quantized variant of Riemannian gradient descent to solve this problem, and
prove that the algorithm converges with high probability under a set of necessary
spherical-convexity properties. We give bounds on the number of bits transmitted
by the algorithm under common initialization schemes, and investigate the dependency
on the problem dimension in each case.
acknowledgement: We would like to thank the anonymous reviewers for helpful comments
and suggestions. We also thank Aurelien Lucchi and Antonio Orvieto for fruitful
discussions at an early stage of this work. FA is partially supported by the SNSF
under research project No. 192363 and conducted part of this work while at IST Austria
under the European Union’s Horizon 2020 research and innovation programme (grant
agreement No. 805223 ScaleML). PD partly conducted this work while at IST Austria
and was supported by the European Union’s Horizon 2020 programme under the Marie
Skłodowska-Curie grant agreement No. 754411.
article_processing_charge: No
arxiv: 1
author:
- first_name: Foivos
full_name: Alimisis, Foivos
last_name: Alimisis
- first_name: Peter
full_name: Davies, Peter
id: 11396234-BB50-11E9-B24C-90FCE5697425
last_name: Davies
orcid: 0000-0002-5646-9524
- first_name: Bart
full_name: Vandereycken, Bart
last_name: Vandereycken
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Alimisis F, Davies P, Vandereycken B, Alistarh D-A. Distributed principal
component analysis with limited communication. In: Advances in Neural Information
Processing Systems - 35th Conference on Neural Information Processing Systems.
Vol 4. Neural Information Processing Systems Foundation; 2021:2823-2834.'
apa: 'Alimisis, F., Davies, P., Vandereycken, B., & Alistarh, D.-A. (2021).
Distributed principal component analysis with limited communication. In Advances
in Neural Information Processing Systems - 35th Conference on Neural Information
Processing Systems (Vol. 4, pp. 2823–2834). Virtual, Online: Neural Information
Processing Systems Foundation.'
chicago: Alimisis, Foivos, Peter Davies, Bart Vandereycken, and Dan-Adrian Alistarh.
“Distributed Principal Component Analysis with Limited Communication.” In Advances
in Neural Information Processing Systems - 35th Conference on Neural Information
Processing Systems, 4:2823–34. Neural Information Processing Systems Foundation,
2021.
ieee: F. Alimisis, P. Davies, B. Vandereycken, and D.-A. Alistarh, “Distributed
principal component analysis with limited communication,” in Advances in Neural
Information Processing Systems - 35th Conference on Neural Information Processing
Systems, Virtual, Online, 2021, vol. 4, pp. 2823–2834.
ista: 'Alimisis F, Davies P, Vandereycken B, Alistarh D-A. 2021. Distributed principal
component analysis with limited communication. Advances in Neural Information
Processing Systems - 35th Conference on Neural Information Processing Systems.
NeurIPS: Neural Information Processing Systems vol. 4, 2823–2834.'
mla: Alimisis, Foivos, et al. “Distributed Principal Component Analysis with Limited
Communication.” Advances in Neural Information Processing Systems - 35th Conference
on Neural Information Processing Systems, vol. 4, Neural Information Processing
Systems Foundation, 2021, pp. 2823–34.
short: F. Alimisis, P. Davies, B. Vandereycken, D.-A. Alistarh, in:, Advances in
Neural Information Processing Systems - 35th Conference on Neural Information
Processing Systems, Neural Information Processing Systems Foundation, 2021, pp.
2823–2834.
conference:
end_date: 2021-12-14
location: Virtual, Online
name: 'NeurIPS: Neural Information Processing Systems'
start_date: 2021-12-06
date_created: 2022-06-19T22:01:58Z
date_published: 2021-12-01T00:00:00Z
date_updated: 2022-06-20T08:31:52Z
day: '01'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2110.14391'
intvolume: ' 4'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://proceedings.neurips.cc/paper/2021/file/1680e9fa7b4dd5d62ece800239bb53bd-Paper.pdf
month: '12'
oa: 1
oa_version: Published Version
page: 2823-2834
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
- _id: 260C2330-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '754411'
name: ISTplus - Postdoctoral Fellowships
publication: Advances in Neural Information Processing Systems - 35th Conference on
Neural Information Processing Systems
publication_identifier:
isbn:
- '9781713845393'
issn:
- 1049-5258
publication_status: published
publisher: Neural Information Processing Systems Foundation
quality_controlled: '1'
scopus_import: '1'
status: public
title: Distributed principal component analysis with limited communication
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 4
year: '2021'
...
---
_id: '11458'
abstract:
- lang: eng
text: 'The increasing computational requirements of deep neural networks (DNNs)
have led to significant interest in obtaining DNN models that are sparse, yet
accurate. Recent work has investigated the even harder case of sparse training,
where the DNN weights are, for as much as possible, already sparse to reduce computational
costs during training. Existing sparse training methods are often empirical and
can have lower accuracy relative to the dense baseline. In this paper, we present
a general approach called Alternating Compressed/DeCompressed (AC/DC) training
of DNNs, demonstrate convergence for a variant of the algorithm, and show that
AC/DC outperforms existing sparse training methods in accuracy at similar computational
budgets; at high sparsity levels, AC/DC even outperforms existing methods that
rely on accurate pre-trained dense models. An important property of AC/DC is that
it allows co-training of dense and sparse models, yielding accurate sparse–dense
model pairs at the end of the training process. This is useful in practice, where
compressed variants may be desirable for deployment in resource-constrained settings
without re-doing the entire training flow, and also provides us with insights
into the accuracy gap between dense and compressed models. The code is available
at: https://github.com/IST-DASLab/ACDC.'
acknowledged_ssus:
- _id: ScienComp
acknowledgement: This project has received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 research and innovation programme
(grant agreement No 805223 ScaleML), and a CNRS PEPS grant. This research was supported
by the Scientific Service Units (SSU) of IST Austria through resources provided
by Scientific Computing (SciComp). We would also like to thank Christoph Lampert
for his feedback on an earlier version of this work, as well as for providing hardware
for the Transformer-XL experiments.
article_processing_charge: No
arxiv: 1
author:
- first_name: Elena-Alexandra
full_name: Peste, Elena-Alexandra
id: 32D78294-F248-11E8-B48F-1D18A9856A87
last_name: Peste
- first_name: Eugenia B
full_name: Iofinova, Eugenia B
id: f9a17499-f6e0-11ea-865d-fdf9a3f77117
last_name: Iofinova
orcid: 0000-0002-7778-3221
- first_name: Adrian
full_name: Vladu, Adrian
last_name: Vladu
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Peste E-A, Iofinova EB, Vladu A, Alistarh D-A. AC/DC: Alternating Compressed/DeCompressed
training of deep neural networks. In: 35th Conference on Neural Information
Processing Systems. Vol 34. Curran Associates; 2021:8557-8570.'
apa: 'Peste, E.-A., Iofinova, E. B., Vladu, A., & Alistarh, D.-A. (2021). AC/DC:
Alternating Compressed/DeCompressed training of deep neural networks. In 35th
Conference on Neural Information Processing Systems (Vol. 34, pp. 8557–8570).
Virtual, Online: Curran Associates.'
chicago: 'Peste, Elena-Alexandra, Eugenia B Iofinova, Adrian Vladu, and Dan-Adrian
Alistarh. “AC/DC: Alternating Compressed/DeCompressed Training of Deep Neural
Networks.” In 35th Conference on Neural Information Processing Systems,
34:8557–70. Curran Associates, 2021.'
ieee: 'E.-A. Peste, E. B. Iofinova, A. Vladu, and D.-A. Alistarh, “AC/DC: Alternating
Compressed/DeCompressed training of deep neural networks,” in 35th Conference
on Neural Information Processing Systems, Virtual, Online, 2021, vol. 34,
pp. 8557–8570.'
ista: 'Peste E-A, Iofinova EB, Vladu A, Alistarh D-A. 2021. AC/DC: Alternating Compressed/DeCompressed
training of deep neural networks. 35th Conference on Neural Information Processing
Systems. NeurIPS: Neural Information Processing Systems vol. 34, 8557–8570.'
mla: 'Peste, Elena-Alexandra, et al. “AC/DC: Alternating Compressed/DeCompressed
Training of Deep Neural Networks.” 35th Conference on Neural Information Processing
Systems, vol. 34, Curran Associates, 2021, pp. 8557–70.'
short: E.-A. Peste, E.B. Iofinova, A. Vladu, D.-A. Alistarh, in:, 35th Conference
on Neural Information Processing Systems, Curran Associates, 2021, pp. 8557–8570.
conference:
end_date: 2021-12-14
location: Virtual, Online
name: 'NeurIPS: Neural Information Processing Systems'
start_date: 2021-12-06
date_created: 2022-06-20T12:11:53Z
date_published: 2021-12-06T00:00:00Z
date_updated: 2023-06-01T12:54:45Z
day: '6'
department:
- _id: GradSch
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2106.12379'
intvolume: ' 34'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://proceedings.neurips.cc/paper/2021/file/48000647b315f6f00f913caa757a70b3-Paper.pdf
month: '12'
oa: 1
oa_version: Published Version
page: 8557-8570
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 35th Conference on Neural Information Processing Systems
publication_identifier:
isbn:
- '9781713845393'
issn:
- 1049-5258
publication_status: published
publisher: Curran Associates
quality_controlled: '1'
related_material:
record:
- id: '13074'
relation: dissertation_contains
status: public
scopus_import: '1'
status: public
title: 'AC/DC: Alternating Compressed/DeCompressed training of deep neural networks'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 34
year: '2021'
...
---
_id: '11463'
abstract:
- lang: eng
text: "Efficiently approximating local curvature information of the loss function
is a key tool for optimization and compression of deep neural networks. Yet, most
existing methods to approximate second-order information have high computational\r\nor
storage costs, which limits their practicality. In this work, we investigate matrix-free,
linear-time approaches for estimating Inverse-Hessian Vector Products (IHVPs)
for the case when the Hessian can be approximated as a sum of rank-one matrices,
as in the classic approximation of the Hessian by the empirical Fisher matrix.
We propose two new algorithms: the first is tailored towards network compression
and can compute the IHVP for dimension d, if the Hessian is given as a sum of
m rank-one matrices, using O(dm2) precomputation, O(dm) cost for computing the
IHVP, and query cost O(m) for any single element of the inverse Hessian. The second
algorithm targets an optimization setting, where we wish to compute the product
between the inverse Hessian, estimated over a sliding window of optimization steps,
and a given gradient direction, as required for preconditioned SGD. We give an
algorithm with cost O(dm + m2) for computing the IHVP and O(dm + m3) for adding
or removing any gradient from the sliding window. These\r\ntwo algorithms yield
state-of-the-art results for network pruning and optimization with lower computational
overhead relative to existing second-order methods. Implementations are available
at [9] and [17]."
acknowledgement: We gratefully acknowledge funding the European Research Council (ERC)
under the European Union’s Horizon 2020 research and innovation programme (grant
agreement No 805223 ScaleML), as well as computational support from Amazon Web Services
(AWS) EC2.
article_processing_charge: No
arxiv: 1
author:
- first_name: Elias
full_name: Frantar, Elias
id: 09a8f98d-ec99-11ea-ae11-c063a7b7fe5f
last_name: Frantar
- first_name: Eldar
full_name: Kurtic, Eldar
id: 47beb3a5-07b5-11eb-9b87-b108ec578218
last_name: Kurtic
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
citation:
ama: 'Frantar E, Kurtic E, Alistarh D-A. M-FAC: Efficient matrix-free approximations
of second-order information. In: 35th Conference on Neural Information Processing
Systems. Vol 34. Curran Associates; 2021:14873-14886.'
apa: 'Frantar, E., Kurtic, E., & Alistarh, D.-A. (2021). M-FAC: Efficient matrix-free
approximations of second-order information. In 35th Conference on Neural Information
Processing Systems (Vol. 34, pp. 14873–14886). Virtual, Online: Curran Associates.'
chicago: 'Frantar, Elias, Eldar Kurtic, and Dan-Adrian Alistarh. “M-FAC: Efficient
Matrix-Free Approximations of Second-Order Information.” In 35th Conference
on Neural Information Processing Systems, 34:14873–86. Curran Associates,
2021.'
ieee: 'E. Frantar, E. Kurtic, and D.-A. Alistarh, “M-FAC: Efficient matrix-free
approximations of second-order information,” in 35th Conference on Neural Information
Processing Systems, Virtual, Online, 2021, vol. 34, pp. 14873–14886.'
ista: 'Frantar E, Kurtic E, Alistarh D-A. 2021. M-FAC: Efficient matrix-free approximations
of second-order information. 35th Conference on Neural Information Processing
Systems. NeurIPS: Neural Information Processing Systems vol. 34, 14873–14886.'
mla: 'Frantar, Elias, et al. “M-FAC: Efficient Matrix-Free Approximations of Second-Order
Information.” 35th Conference on Neural Information Processing Systems,
vol. 34, Curran Associates, 2021, pp. 14873–86.'
short: E. Frantar, E. Kurtic, D.-A. Alistarh, in:, 35th Conference on Neural Information
Processing Systems, Curran Associates, 2021, pp. 14873–14886.
conference:
end_date: 2021-12-14
location: Virtual, Online
name: 'NeurIPS: Neural Information Processing Systems'
start_date: 2021-12-06
date_created: 2022-06-26T22:01:35Z
date_published: 2021-12-06T00:00:00Z
date_updated: 2022-06-27T07:05:12Z
day: '06'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2010.08222'
intvolume: ' 34'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://proceedings.neurips.cc/paper/2021/file/7cfd5df443b4eb0d69886a583b33de4c-Paper.pdf
month: '12'
oa: 1
oa_version: Published Version
page: 14873-14886
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 35th Conference on Neural Information Processing Systems
publication_identifier:
isbn:
- '9781713845393'
issn:
- 1049-5258
publication_status: published
publisher: Curran Associates
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'M-FAC: Efficient matrix-free approximations of second-order information'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 34
year: '2021'
...
---
_id: '11464'
abstract:
- lang: eng
text: "We consider a standard distributed optimisation setting where N machines,
each holding a d-dimensional function\r\nfi, aim to jointly minimise the sum of
the functions ∑Ni=1fi(x). This problem arises naturally in large-scale distributed
optimisation, where a standard solution is to apply variants of (stochastic) gradient
descent. We focus on the communication complexity of this problem: our main result
provides the first fully unconditional bounds on total number of bits which need
to be sent and received by the N machines to solve this problem under point-to-point
communication, within a given error-tolerance. Specifically, we show that Ω(Ndlogd/Nε)
total bits need to be communicated between the machines to find an additive ϵ-approximation
to the minimum of ∑Ni=1fi(x). The result holds for both deterministic and randomised
algorithms, and, importantly, requires no assumptions on the algorithm structure.
The lower bound is tight under certain restrictions on parameter values, and is
matched within constant factors for quadratic objectives by a new variant of quantised
gradient descent, which we describe and analyse. Our results bring over tools
from communication complexity to distributed optimisation, which has potential
for further applications."
acknowledgement: We thank the NeurIPS reviewers for insightful comments that helped
us improve the positioning of our results, as well as for pointing out the subsampling
approach for complementing the randomised lower bound. We also thank Foivos Alimisis
and Peter Davies for useful discussions. This project has received funding from
the European Research Council (ERC) under the European Union’s Horizon 2020 research
and innovation programme (grant agreement No 805223 ScaleML).
article_processing_charge: No
arxiv: 1
author:
- first_name: Dan-Adrian
full_name: Alistarh, Dan-Adrian
id: 4A899BFC-F248-11E8-B48F-1D18A9856A87
last_name: Alistarh
orcid: 0000-0003-3650-940X
- first_name: Janne
full_name: Korhonen, Janne
id: C5402D42-15BC-11E9-A202-CA2BE6697425
last_name: Korhonen
citation:
ama: 'Alistarh D-A, Korhonen J. Towards tight communication lower bounds for distributed
optimisation. In: 35th Conference on Neural Information Processing Systems.
Vol 34. Curran Associates; 2021:7254-7266.'
apa: 'Alistarh, D.-A., & Korhonen, J. (2021). Towards tight communication lower
bounds for distributed optimisation. In 35th Conference on Neural Information
Processing Systems (Vol. 34, pp. 7254–7266). Virtual, Online: Curran Associates.'
chicago: Alistarh, Dan-Adrian, and Janne Korhonen. “Towards Tight Communication
Lower Bounds for Distributed Optimisation.” In 35th Conference on Neural Information
Processing Systems, 34:7254–66. Curran Associates, 2021.
ieee: D.-A. Alistarh and J. Korhonen, “Towards tight communication lower bounds
for distributed optimisation,” in 35th Conference on Neural Information Processing
Systems, Virtual, Online, 2021, vol. 34, pp. 7254–7266.
ista: 'Alistarh D-A, Korhonen J. 2021. Towards tight communication lower bounds
for distributed optimisation. 35th Conference on Neural Information Processing
Systems. NeurIPS: Neural Information Processing Systems vol. 34, 7254–7266.'
mla: Alistarh, Dan-Adrian, and Janne Korhonen. “Towards Tight Communication Lower
Bounds for Distributed Optimisation.” 35th Conference on Neural Information
Processing Systems, vol. 34, Curran Associates, 2021, pp. 7254–66.
short: D.-A. Alistarh, J. Korhonen, in:, 35th Conference on Neural Information Processing
Systems, Curran Associates, 2021, pp. 7254–7266.
conference:
end_date: 2021-12-14
location: Virtual, Online
name: 'NeurIPS: Neural Information Processing Systems'
start_date: 2021-12-06
date_created: 2022-06-26T22:01:35Z
date_published: 2021-12-06T00:00:00Z
date_updated: 2022-06-27T06:54:31Z
day: '06'
department:
- _id: DaAl
ec_funded: 1
external_id:
arxiv:
- '2010.08222'
intvolume: ' 34'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://proceedings.neurips.cc/paper/2021/file/3b92d18aa7a6176dd37d372bc2f1eb71-Paper.pdf
month: '12'
oa: 1
oa_version: Published Version
page: 7254-7266
project:
- _id: 268A44D6-B435-11E9-9278-68D0E5697425
call_identifier: H2020
grant_number: '805223'
name: Elastic Coordination for Scalable Machine Learning
publication: 35th Conference on Neural Information Processing Systems
publication_identifier:
isbn:
- '9781713845393'
issn:
- 1049-5258
publication_status: published
publisher: Curran Associates
quality_controlled: '1'
scopus_import: '1'
status: public
title: Towards tight communication lower bounds for distributed optimisation
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 34
year: '2021'
...