---
_id: '14213'
abstract:
- lang: eng
text: We introduce a method to segment the visual field into independently moving
regions, trained with no ground truth or supervision. It consists of an adversarial
conditional encoder-decoder architecture based on Slot Attention, modified to
use the image as context to decode optical flow without attempting to reconstruct
the image itself. In the resulting multi-modal representation, one modality (flow)
feeds the encoder to produce separate latent codes (slots), whereas the other
modality (image) conditions the decoder to generate the first (flow) from the
slots. This design frees the representation from having to encode complex nuisance
variability in the image due to, for instance, illumination and reflectance properties
of the scene. Since customary autoencoding based on minimizing the reconstruction
error does not preclude the entire flow from being encoded into a single slot,
we modify the loss to an adversarial criterion based on Contextual Information
Separation. The resulting min-max optimization fosters the separation of objects
and their assignment to different attention slots, leading to Divided Attention,
or DivA. DivA outperforms recent unsupervised multi-object motion segmentation
methods while tripling run-time speed up to 104FPS and reducing the performance
gap from supervised methods to 12% or less. DivA can handle different numbers
of objects and different image sizes at training and test time, is invariant to
permutation of object labels, and does not require explicit regularization.
article_processing_charge: No
arxiv: 1
author:
- first_name: Dong
full_name: Lao, Dong
last_name: Lao
- first_name: Zhengyang
full_name: Hu, Zhengyang
last_name: Hu
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Yanchao
full_name: Yang, Yanchao
last_name: Yang
- first_name: Stefano
full_name: Soatto, Stefano
last_name: Soatto
conference:
end_date: 2024-01-03
location: Hong Kong, China
name: 'CPAL: Conference on Parsimony and Learning'
start_date: 2024-01-03
date_created: 2023-08-22T14:19:59Z
date_published: 2024-01-03T00:00:00Z
date_updated: 2025-02-13T08:10:28Z
day: '03'
ddc:
- '000'
department:
- _id: FrLo
external_id:
arxiv:
- '2304.01430'
file:
- access_level: open_access
checksum: 8fad894c34f1b3d5a14fb8ffb12f7277
content_type: application/pdf
creator: dernst
date_created: 2024-02-12T08:40:36Z
date_updated: 2024-02-12T08:40:36Z
file_id: '14978'
file_name: 2024_CPAL_Lao.pdf
file_size: 8038511
relation: main_file
success: 1
file_date_updated: 2024-02-12T08:40:36Z
has_accepted_license: '1'
language:
- iso: eng
month: '01'
oa: 1
oa_version: Published Version
publication: 1st Conference on Parsimony and Learning
publication_identifier:
issnl:
- 1234-4321
publication_status: published
quality_controlled: '1'
status: public
title: 'Divided attention: Unsupervised multi-object discovery with contextually separated
slots'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2024'
...
---
_id: '14333'
abstract:
- lang: eng
text: "As causal ground truth is incredibly rare, causal discovery algorithms are\r\ncommonly
only evaluated on simulated data. This is concerning, given that\r\nsimulations
reflect common preconceptions about generating processes regarding\r\nnoise distributions,
model classes, and more. In this work, we propose a novel\r\nmethod for falsifying
the output of a causal discovery algorithm in the absence\r\nof ground truth.
Our key insight is that while statistical learning seeks\r\nstability across subsets
of data points, causal learning should seek stability\r\nacross subsets of variables.
Motivated by this insight, our method relies on a\r\nnotion of compatibility between
causal graphs learned on different subsets of\r\nvariables. We prove that detecting
incompatibilities can falsify wrongly\r\ninferred causal relations due to violation
of assumptions or errors from finite\r\nsample effects. Although passing such
compatibility tests is only a necessary\r\ncriterion for good performance, we
argue that it provides strong evidence for\r\nthe causal models whenever compatibility
entails strong implications for the\r\njoint distribution. We also demonstrate
experimentally that detection of\r\nincompatibilities can aid in causal model
selection."
article_number: '2307.09552'
article_processing_charge: No
arxiv: 1
author:
- first_name: Philipp M.
full_name: Faller, Philipp M.
last_name: Faller
- first_name: Leena Chennuru
full_name: Vankadara, Leena Chennuru
last_name: Vankadara
- first_name: Atalanti A.
full_name: Mastakouri, Atalanti A.
last_name: Mastakouri
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Dominik
full_name: Janzing, Dominik
last_name: Janzing
citation:
ama: 'Faller PM, Vankadara LC, Mastakouri AA, Locatello F, Janzing D. Self-compatibility:
Evaluating causal discovery without ground truth. arXiv. doi:10.48550/arXiv.2307.09552'
apa: 'Faller, P. M., Vankadara, L. C., Mastakouri, A. A., Locatello, F., & Janzing,
D. (n.d.). Self-compatibility: Evaluating causal discovery without ground truth.
arXiv. https://doi.org/10.48550/arXiv.2307.09552'
chicago: 'Faller, Philipp M., Leena Chennuru Vankadara, Atalanti A. Mastakouri,
Francesco Locatello, and Dominik Janzing. “Self-Compatibility: Evaluating Causal
Discovery without Ground Truth.” ArXiv, n.d. https://doi.org/10.48550/arXiv.2307.09552.'
ieee: 'P. M. Faller, L. C. Vankadara, A. A. Mastakouri, F. Locatello, and D. Janzing,
“Self-compatibility: Evaluating causal discovery without ground truth,” arXiv.
.'
ista: 'Faller PM, Vankadara LC, Mastakouri AA, Locatello F, Janzing D. Self-compatibility:
Evaluating causal discovery without ground truth. arXiv, 2307.09552.'
mla: 'Faller, Philipp M., et al. “Self-Compatibility: Evaluating Causal Discovery
without Ground Truth.” ArXiv, 2307.09552, doi:10.48550/arXiv.2307.09552.'
short: P.M. Faller, L.C. Vankadara, A.A. Mastakouri, F. Locatello, D. Janzing, ArXiv
(n.d.).
date_created: 2023-09-13T12:44:59Z
date_published: 2023-07-18T00:00:00Z
date_updated: 2023-09-13T12:47:53Z
day: '18'
department:
- _id: FrLo
doi: 10.48550/arXiv.2307.09552
extern: '1'
external_id:
arxiv:
- '2307.09552'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2307.09552
month: '07'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: 'Self-compatibility: Evaluating causal discovery without ground truth'
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14946'
abstract:
- lang: eng
text: "We present a unified framework for studying the identifiability of\r\nrepresentations
learned from simultaneously observed views, such as different\r\ndata modalities.
We allow a partially observed setting in which each view\r\nconstitutes a nonlinear
mixture of a subset of underlying latent variables,\r\nwhich can be causally related.
We prove that the information shared across all\r\nsubsets of any number of views
can be learned up to a smooth bijection using\r\ncontrastive learning and a single
encoder per view. We also provide graphical\r\ncriteria indicating which latent
variables can be identified through a simple\r\nset of rules, which we refer to
as identifiability algebra. Our general\r\nframework and theoretical results unify
and extend several previous works on\r\nmulti-view nonlinear ICA, disentanglement,
and causal representation learning.\r\nWe experimentally validate our claims on
numerical, image, and multi-modal data\r\nsets. Further, we demonstrate that the
performance of prior methods is\r\nrecovered in different special cases of our
setup. Overall, we find that access\r\nto multiple partial views enables us to
identify a more fine-grained\r\nrepresentation, under the generally milder assumption
of partial observability."
acknowledgement: "This work was initiated at the Second Bellairs Workshop on Causality
held at the Bellairs Research Institute, January 6–13, 2022; we thank all workshop
participants for providing a stimulating research environment. Further, we thank
Cian Eastwood, Luigi Gresele, Stefano Soatto, Marco Bagatella, and A. René Geist
for helpful discussion. GM is a member of the Machine Learning Cluster of Excellence,
EXC number 2064/1 – Project number 390727645. JvK and GM acknowledge support from
the German Federal Ministry of Education and Research (BMBF) through the Tübingen
AI Center (FKZ: 01IS18039B). The research of DX and SM was supported by the Air
Force Office of Scientific Research under award number FA8655-22-1-7155. Any opinions,
findings, and conclusions or recommendations expressed in\r\nthis material are those
of the author(s) and do not necessarily reflect the views of the United States Air
Force. We also thank SURF for the support in using the Dutch National Supercomputer
Snellius. DY was supported by an Amazon fellowship and the International Max Planck
Research School for Intelligent Systems (IMPRS-IS). Work done outside of Amazon.
SL was supported by an IVADO excellence PhD scholarship and by Samsung Electronics
Co., Ldt."
article_number: '2311.04056'
article_processing_charge: No
arxiv: 1
author:
- first_name: Dingling
full_name: Yao, Dingling
id: d3e02e50-48a8-11ee-8f62-c108061797fa
last_name: Yao
- first_name: Danru
full_name: Xu, Danru
last_name: Xu
- first_name: Sébastien
full_name: Lachapelle, Sébastien
last_name: Lachapelle
- first_name: Sara
full_name: Magliacane, Sara
last_name: Magliacane
- first_name: Perouz
full_name: Taslakian, Perouz
last_name: Taslakian
- first_name: Georg
full_name: Martius, Georg
last_name: Martius
- first_name: Julius von
full_name: Kügelgen, Julius von
last_name: Kügelgen
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: Yao D, Xu D, Lachapelle S, et al. Multi-view causal representation learning
with partial observability. arXiv. doi:10.48550/arXiv.2311.04056
apa: Yao, D., Xu, D., Lachapelle, S., Magliacane, S., Taslakian, P., Martius, G.,
… Locatello, F. (n.d.). Multi-view causal representation learning with partial
observability. arXiv. https://doi.org/10.48550/arXiv.2311.04056
chicago: Yao, Dingling, Danru Xu, Sébastien Lachapelle, Sara Magliacane, Perouz
Taslakian, Georg Martius, Julius von Kügelgen, and Francesco Locatello. “Multi-View
Causal Representation Learning with Partial Observability.” ArXiv, n.d.
https://doi.org/10.48550/arXiv.2311.04056.
ieee: D. Yao et al., “Multi-view causal representation learning with partial
observability,” arXiv. .
ista: Yao D, Xu D, Lachapelle S, Magliacane S, Taslakian P, Martius G, Kügelgen
J von, Locatello F. Multi-view causal representation learning with partial observability.
arXiv, 2311.04056.
mla: Yao, Dingling, et al. “Multi-View Causal Representation Learning with Partial
Observability.” ArXiv, 2311.04056, doi:10.48550/arXiv.2311.04056.
short: D. Yao, D. Xu, S. Lachapelle, S. Magliacane, P. Taslakian, G. Martius, J.
von Kügelgen, F. Locatello, ArXiv (n.d.).
date_created: 2024-02-07T14:28:34Z
date_published: 2023-11-07T00:00:00Z
date_updated: 2024-02-12T08:07:33Z
day: '07'
department:
- _id: FrLo
doi: 10.48550/arXiv.2311.04056
external_id:
arxiv:
- '2311.04056'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2311.04056
month: '11'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Multi-view causal representation learning with partial observability
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14948'
abstract:
- lang: eng
text: "The extraction of modular object-centric representations for downstream tasks\r\nis
an emerging area of research. Learning grounded representations of objects\r\nthat
are guaranteed to be stable and invariant promises robust performance\r\nacross
different tasks and environments. Slot Attention (SA) learns\r\nobject-centric
representations by assigning objects to \\textit{slots}, but\r\npresupposes a
\\textit{single} distribution from which all slots are randomly\r\ninitialised.
This results in an inability to learn \\textit{specialized} slots\r\nwhich bind
to specific object types and remain invariant to identity-preserving\r\nchanges
in object appearance. To address this, we present\r\n\\emph{\\textsc{Co}nditional
\\textsc{S}lot \\textsc{A}ttention} (\\textsc{CoSA})\r\nusing a novel concept
of \\emph{Grounded Slot Dictionary} (GSD) inspired by\r\nvector quantization.
Our proposed GSD comprises (i) canonical object-level\r\nproperty vectors and
(ii) parametric Gaussian distributions, which define a\r\nprior over the slots.
We demonstrate the benefits of our method in multiple\r\ndownstream tasks such
as scene generation, composition, and task adaptation,\r\nwhilst remaining competitive
with SA in popular object discovery benchmarks."
acknowledgement: "This work was supported by supported by UKRI (grant agreement no.
EP/S023356/1), in the UKRI\r\nCentre for Doctoral Training in Safe and Trusted AI
via A. Kori."
article_number: '2307.09437'
article_processing_charge: No
arxiv: 1
author:
- first_name: Avinash
full_name: Kori, Avinash
last_name: Kori
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Fabio De Sousa
full_name: Ribeiro, Fabio De Sousa
last_name: Ribeiro
- first_name: Francesca
full_name: Toni, Francesca
last_name: Toni
- first_name: Ben
full_name: Glocker, Ben
last_name: Glocker
citation:
ama: Kori A, Locatello F, Ribeiro FDS, Toni F, Glocker B. Grounded object centric
learning. arXiv. doi:10.48550/arXiv.2307.09437
apa: Kori, A., Locatello, F., Ribeiro, F. D. S., Toni, F., & Glocker, B. (n.d.).
Grounded object centric learning. arXiv. https://doi.org/10.48550/arXiv.2307.09437
chicago: Kori, Avinash, Francesco Locatello, Fabio De Sousa Ribeiro, Francesca Toni,
and Ben Glocker. “Grounded Object Centric Learning.” ArXiv, n.d. https://doi.org/10.48550/arXiv.2307.09437.
ieee: A. Kori, F. Locatello, F. D. S. Ribeiro, F. Toni, and B. Glocker, “Grounded
object centric learning,” arXiv. .
ista: Kori A, Locatello F, Ribeiro FDS, Toni F, Glocker B. Grounded object centric
learning. arXiv, 2307.09437.
mla: Kori, Avinash, et al. “Grounded Object Centric Learning.” ArXiv, 2307.09437,
doi:10.48550/arXiv.2307.09437.
short: A. Kori, F. Locatello, F.D.S. Ribeiro, F. Toni, B. Glocker, ArXiv (n.d.).
date_created: 2024-02-07T14:47:04Z
date_published: 2023-07-18T00:00:00Z
date_updated: 2024-02-12T08:13:12Z
day: '18'
department:
- _id: FrLo
doi: 10.48550/arXiv.2307.09437
external_id:
arxiv:
- '2307.09437'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2307.09437
month: '07'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Grounded object centric learning
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14949'
abstract:
- lang: eng
text: Many approaches have been proposed to use diffusion models to augment training
datasets for downstream tasks, such as classification. However, diffusion models
are themselves trained on large datasets, often with noisy annotations, and it
remains an open question to which extent these models contribute to downstream
classification performance. In particular, it remains unclear if they generalize
enough to improve over directly using the additional data of their pre-training
process for augmentation. We systematically evaluate a range of existing methods
to generate images from diffusion models and study new extensions to assess their
benefit for data augmentation. Personalizing diffusion models towards the target
data outperforms simpler prompting strategies. However, using the pre-training
data of the diffusion model alone, via a simple nearest-neighbor retrieval procedure,
leads to even stronger downstream performance. Our study explores the potential
of diffusion models in generating new training data, and surprisingly finds that
these sophisticated models are not yet able to beat a simple and strong image
retrieval baseline on simple downstream vision tasks.
acknowledgement: The authors would like to thank Varad Gunjal and Vishaal Udandarao.
MFB thanks the International Max Planck Research School for Intelligent Systems
(IMPRS-IS).
alternative_title:
- TMLR
article_processing_charge: No
article_type: original
author:
- first_name: Max
full_name: Burg, Max
last_name: Burg
- first_name: Florian
full_name: Wenzel, Florian
last_name: Wenzel
- first_name: Dominik
full_name: Zietlow, Dominik
last_name: Zietlow
- first_name: Max
full_name: Horn, Max
last_name: Horn
- first_name: Osama
full_name: Makansi, Osama
last_name: Makansi
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Chris
full_name: Russell, Chris
last_name: Russell
citation:
ama: Burg M, Wenzel F, Zietlow D, et al. Image retrieval outperforms diffusion models
on data augmentation. Journal of Machine Learning Research. 2023.
apa: Burg, M., Wenzel, F., Zietlow, D., Horn, M., Makansi, O., Locatello, F., &
Russell, C. (2023). Image retrieval outperforms diffusion models on data augmentation.
Journal of Machine Learning Research. ML Research Press.
chicago: Burg, Max, Florian Wenzel, Dominik Zietlow, Max Horn, Osama Makansi, Francesco
Locatello, and Chris Russell. “Image Retrieval Outperforms Diffusion Models on
Data Augmentation.” Journal of Machine Learning Research. ML Research Press,
2023.
ieee: M. Burg et al., “Image retrieval outperforms diffusion models on data
augmentation,” Journal of Machine Learning Research. ML Research Press,
2023.
ista: Burg M, Wenzel F, Zietlow D, Horn M, Makansi O, Locatello F, Russell C. 2023.
Image retrieval outperforms diffusion models on data augmentation. Journal of
Machine Learning Research.
mla: Burg, Max, et al. “Image Retrieval Outperforms Diffusion Models on Data Augmentation.”
Journal of Machine Learning Research, ML Research Press, 2023.
short: M. Burg, F. Wenzel, D. Zietlow, M. Horn, O. Makansi, F. Locatello, C. Russell,
Journal of Machine Learning Research (2023).
date_created: 2024-02-07T14:57:39Z
date_published: 2023-12-10T00:00:00Z
date_updated: 2024-02-12T08:30:21Z
day: '10'
ddc:
- '000'
department:
- _id: FrLo
file:
- access_level: open_access
checksum: af87ddea7908923426365347b9c87ba7
content_type: application/pdf
creator: ptazenko
date_created: 2024-02-07T14:57:32Z
date_updated: 2024-02-07T14:57:32Z
file_id: '14950'
file_name: Burg_et_al_2023_Image_retrieval_outperforms.pdf
file_size: 27325153
relation: main_file
file_date_updated: 2024-02-07T14:57:32Z
has_accepted_license: '1'
language:
- iso: eng
license: https://creativecommons.org/licenses/by/4.0/
main_file_link:
- open_access: '1'
url: https://openreview.net/forum?id=xflYdGZMpv
month: '12'
oa: 1
oa_version: Published Version
publication: Journal of Machine Learning Research
publication_identifier:
eissn:
- 2835-8856
publication_status: published
publisher: ML Research Press
quality_controlled: '1'
status: public
title: Image retrieval outperforms diffusion models on data augmentation
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: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14952'
abstract:
- lang: eng
text: "While different neural models often exhibit latent spaces that are alike
when exposed to semantically related data, this intrinsic similarity is not always
immediately discernible. Towards a better understanding of this phenomenon, our
work shows how representations learned from these neural modules can be translated
between different pre-trained networks via simpler transformations than previously
thought. An advantage of this approach is the ability to\r\nestimate these transformations
using standard, well-understood algebraic procedures that have closed-form solutions.
Our method directly estimates a transformation between two given latent spaces,
thereby enabling effective stitching of encoders and decoders without additional
training. We extensively validate the adaptability of this translation procedure
in different\r\nexperimental settings: across various trainings, domains, architectures
(e.g., ResNet, CNN, ViT), and in multiple downstream tasks (classification, reconstruction).
Notably, we show how it is possible to zero-shot stitch text encoders and vision
decoders, or vice-versa, yielding surprisingly good classification performance
in this multimodal setting."
acknowledgement: "This work is supported by the ERC grant no.802554 (SPECGEO), PRIN
2020 project no.2020TA3K9N (LEGO.AI), and PNRR MUR project PE0000013-FAIR. Francesco\r\nLocatello
did not contribute to this work at Amazon."
article_number: '2311.00664'
article_processing_charge: No
arxiv: 1
author:
- first_name: Valentino
full_name: Maiorca, Valentino
last_name: Maiorca
- first_name: Luca
full_name: Moschella, Luca
last_name: Moschella
- first_name: Antonio
full_name: Norelli, Antonio
last_name: Norelli
- first_name: Marco
full_name: Fumero, Marco
last_name: Fumero
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Emanuele
full_name: Rodolà, Emanuele
last_name: Rodolà
citation:
ama: Maiorca V, Moschella L, Norelli A, Fumero M, Locatello F, Rodolà E. Latent
space translation via semantic alignment. arXiv. doi:10.48550/arXiv.2311.00664
apa: Maiorca, V., Moschella, L., Norelli, A., Fumero, M., Locatello, F., & Rodolà,
E. (n.d.). Latent space translation via semantic alignment. arXiv. https://doi.org/10.48550/arXiv.2311.00664
chicago: Maiorca, Valentino, Luca Moschella, Antonio Norelli, Marco Fumero, Francesco
Locatello, and Emanuele Rodolà. “Latent Space Translation via Semantic Alignment.”
ArXiv, n.d. https://doi.org/10.48550/arXiv.2311.00664.
ieee: V. Maiorca, L. Moschella, A. Norelli, M. Fumero, F. Locatello, and E. Rodolà,
“Latent space translation via semantic alignment,” arXiv. .
ista: Maiorca V, Moschella L, Norelli A, Fumero M, Locatello F, Rodolà E. Latent
space translation via semantic alignment. arXiv, 2311.00664.
mla: Maiorca, Valentino, et al. “Latent Space Translation via Semantic Alignment.”
ArXiv, 2311.00664, doi:10.48550/arXiv.2311.00664.
short: V. Maiorca, L. Moschella, A. Norelli, M. Fumero, F. Locatello, E. Rodolà,
ArXiv (n.d.).
date_created: 2024-02-07T15:08:55Z
date_published: 2023-11-01T00:00:00Z
date_updated: 2024-02-12T09:40:23Z
day: '01'
department:
- _id: FrLo
doi: 10.48550/arXiv.2311.00664
external_id:
arxiv:
- '2311.00664'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2311.00664
month: '11'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Latent space translation via semantic alignment
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14953'
abstract:
- lang: eng
text: This paper provides statistical sample complexity bounds for score-matching
and its applications in causal discovery. We demonstrate that accurate estimation
of the score function is achievable by training a standard deep ReLU neural network
using stochastic gradient descent. We establish bounds on the error rate of recovering
causal relationships using the score-matching-based causal discovery method of
Rolland et al. [2022], assuming a sufficiently good estimation of the score function.
Finally, we analyze the upper bound of score-matching estimation within the score-based
generative modeling, which has been applied for causal discovery but is also of
independent interest within the domain of generative models.
acknowledgement: 'We are thankful to the reviewers for providing constructive feedback
and Kun Zhang and Dominik Janzing for helpful discussion on the special case of
deterministic children. This work was supported by Hasler Foundation Program: Hasler
Responsible AI (project number 21043). This work was supported by the Swiss National
Science Foundation (SNSF) under grant number 200021_205011. Francesco Locatello
did not contribute to this work at Amazon. '
article_number: '2310.18123'
article_processing_charge: No
arxiv: 1
author:
- first_name: Zhenyu
full_name: Zhu, Zhenyu
last_name: Zhu
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Volkan
full_name: Cevher, Volkan
last_name: Cevher
citation:
ama: 'Zhu Z, Locatello F, Cevher V. Sample complexity bounds for score-matching:
Causal discovery and generative modeling. arXiv. doi:10.48550/arXiv.2310.18123'
apa: 'Zhu, Z., Locatello, F., & Cevher, V. (n.d.). Sample complexity bounds
for score-matching: Causal discovery and generative modeling. arXiv. https://doi.org/10.48550/arXiv.2310.18123'
chicago: 'Zhu, Zhenyu, Francesco Locatello, and Volkan Cevher. “Sample Complexity
Bounds for Score-Matching: Causal Discovery and Generative Modeling.” ArXiv,
n.d. https://doi.org/10.48550/arXiv.2310.18123.'
ieee: 'Z. Zhu, F. Locatello, and V. Cevher, “Sample complexity bounds for score-matching:
Causal discovery and generative modeling,” arXiv. .'
ista: 'Zhu Z, Locatello F, Cevher V. Sample complexity bounds for score-matching:
Causal discovery and generative modeling. arXiv, 2310.18123.'
mla: 'Zhu, Zhenyu, et al. “Sample Complexity Bounds for Score-Matching: Causal Discovery
and Generative Modeling.” ArXiv, 2310.18123, doi:10.48550/arXiv.2310.18123.'
short: Z. Zhu, F. Locatello, V. Cevher, ArXiv (n.d.).
date_created: 2024-02-07T15:11:11Z
date_published: 2023-10-27T00:00:00Z
date_updated: 2024-02-12T09:45:58Z
day: '27'
department:
- _id: FrLo
doi: 10.48550/arXiv.2310.18123
external_id:
arxiv:
- '2310.18123'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2310.18123
month: '10'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: 'Sample complexity bounds for score-matching: Causal discovery and generative
modeling'
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14954'
abstract:
- lang: eng
text: "When domain knowledge is limited and experimentation is restricted by ethical,
financial, or time constraints, practitioners turn to observational causal discovery
methods to recover the causal structure, exploiting the statistical properties
of their data. Because causal discovery without further assumptions is an ill-posed
problem, each algorithm comes with its own set of\r\nusually untestable assumptions,
some of which are hard to meet in real datasets. Motivated by these considerations,
this paper extensively benchmarks the empirical performance of recent causal discovery
methods on observational i.i.d. data generated under different background conditions,
allowing for violations of the critical assumptions required by each selected
approach. Our experimental findings show that score matching-based methods demonstrate\r\nsurprising
performance in the false positive and false negative rate of the inferred graph
in these challenging scenarios, and we provide theoretical insights into their
performance. This work is also the first effort to benchmark the stability of
causal discovery algorithms with respect to the values of their hyperparameters.
Finally, we hope this paper will set a new standard for the evaluation of causal
discovery methods and can serve as an accessible entry point for practitioners
interested in the field, highlighting the empirical implications of different
algorithm choices."
acknowledgement: "We thank Kun Zhang and Carl-Johann Simon-Gabriel for the insightful
discussions. This work\r\nhas been supported by AFOSR, grant n. FA8655-20-1-7035.
FM is supported by Programma\r\nOperativo Nazionale ricerca e innovazione 2014-2020.
FM partially contributed to this work during an internship at Amazon Web Services
with FL. FL partially contributed while at AWS."
article_number: '2310.13387'
article_processing_charge: No
arxiv: 1
author:
- first_name: Francesco
full_name: Montagna, Francesco
last_name: Montagna
- first_name: Atalanti A.
full_name: Mastakouri, Atalanti A.
last_name: Mastakouri
- first_name: Elias
full_name: Eulig, Elias
last_name: Eulig
- first_name: Nicoletta
full_name: Noceti, Nicoletta
last_name: Noceti
- first_name: Lorenzo
full_name: Rosasco, Lorenzo
last_name: Rosasco
- first_name: Dominik
full_name: Janzing, Dominik
last_name: Janzing
- first_name: Bryon
full_name: Aragam, Bryon
last_name: Aragam
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: Montagna F, Mastakouri AA, Eulig E, et al. Assumption violations in causal
discovery and the robustness of score matching. arXiv. doi:10.48550/arXiv.2310.13387
apa: Montagna, F., Mastakouri, A. A., Eulig, E., Noceti, N., Rosasco, L., Janzing,
D., … Locatello, F. (n.d.). Assumption violations in causal discovery and the
robustness of score matching. arXiv. https://doi.org/10.48550/arXiv.2310.13387
chicago: Montagna, Francesco, Atalanti A. Mastakouri, Elias Eulig, Nicoletta Noceti,
Lorenzo Rosasco, Dominik Janzing, Bryon Aragam, and Francesco Locatello. “Assumption
Violations in Causal Discovery and the Robustness of Score Matching.” ArXiv,
n.d. https://doi.org/10.48550/arXiv.2310.13387.
ieee: F. Montagna et al., “Assumption violations in causal discovery and
the robustness of score matching,” arXiv. .
ista: Montagna F, Mastakouri AA, Eulig E, Noceti N, Rosasco L, Janzing D, Aragam
B, Locatello F. Assumption violations in causal discovery and the robustness of
score matching. arXiv, 2310.13387.
mla: Montagna, Francesco, et al. “Assumption Violations in Causal Discovery and
the Robustness of Score Matching.” ArXiv, 2310.13387, doi:10.48550/arXiv.2310.13387.
short: F. Montagna, A.A. Mastakouri, E. Eulig, N. Noceti, L. Rosasco, D. Janzing,
B. Aragam, F. Locatello, ArXiv (n.d.).
date_created: 2024-02-07T15:11:56Z
date_published: 2023-10-20T00:00:00Z
date_updated: 2024-02-12T09:51:15Z
day: '20'
department:
- _id: FrLo
doi: 10.48550/arXiv.2310.13387
external_id:
arxiv:
- '2310.13387'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2310.13387
month: '10'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Assumption violations in causal discovery and the robustness of score matching
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14958'
abstract:
- lang: eng
text: Causal representation learning (CRL) aims at identifying high-level causal
variables from low-level data, e.g. images. Current methods usually assume that
all causal variables are captured in the high-dimensional observations. In this
work, we focus on learning causal representations from data under partial observability,
i.e., when some of the causal variables are not observed in the measurements,
and the set of masked variables changes across the different samples. We introduce
some initial theoretical results for identifying causal variables under partial
observability by exploiting a sparsity regularizer, focusing in particular on
the linear and piecewise linear mixing function case. We provide a theorem that
allows us to identify the causal variables up to permutation and element-wise
linear transformations in the linear case and a lemma that allows us to identify
causal variables up to linear transformation in the piecewise case. Finally, we
provide a conjecture that would allow us to identify the causal variables up to
permutation and element-wise linear transformations also in the piecewise linear
case. We test the theorem and conjecture on simulated data, showing the effectiveness
of our method.
acknowledgement: "This work was initiated at the Second Bellairs Workshop on Causality
held at the Bellairs Research Institute, January 6–13, 2022; we thank all workshop
participants for providing a stimulating research environment. The research of DX
and SM was supported by the Air Force Office of Scientific Research under award
number FA8655-22-1-7155. Any opinions, findings, and conclusions or recommendations
expressed in this material are those of the author(s) and do not necessarily reflect
the views of the United States Air Force. We also thank SURF for the support in
using the Dutch National Supercomputer Snellius. DY was supported by an Amazon fellowship
and the International Max Planck Research School for Intelligent Systems (IMPRS-IS).
Work done outside of Amazon. SL was supported by an IVADO excellence PhD scholarship
and by Samsung Electronics Co., Ldt. JvK acknowledges support from the German Federal
Ministry of Education and Research (BMBF)\r\nthrough the Tübingen AI Center (FKZ:
01IS18039B).\r\n"
article_number: '54'
article_processing_charge: No
author:
- first_name: Danru
full_name: Xu, Danru
last_name: Xu
- first_name: Dingling
full_name: Yao, Dingling
id: d3e02e50-48a8-11ee-8f62-c108061797fa
last_name: Yao
- first_name: Sebastien
full_name: Lachapelle, Sebastien
last_name: Lachapelle
- first_name: Perouz
full_name: Taslakian, Perouz
last_name: Taslakian
- first_name: Julius
full_name: von Kügelgen, Julius
last_name: von Kügelgen
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Sara
full_name: Magliacane, Sara
last_name: Magliacane
citation:
ama: 'Xu D, Yao D, Lachapelle S, et al. A sparsity principle for partially observable
causal representation learning. In: Causal Representation Learning Workshop
at NeurIPS 2023. OpenReview; 2023.'
apa: 'Xu, D., Yao, D., Lachapelle, S., Taslakian, P., von Kügelgen, J., Locatello,
F., & Magliacane, S. (2023). A sparsity principle for partially observable
causal representation learning. In Causal Representation Learning Workshop
at NeurIPS 2023. New Orleans, LA, United States: OpenReview.'
chicago: Xu, Danru, Dingling Yao, Sebastien Lachapelle, Perouz Taslakian, Julius
von Kügelgen, Francesco Locatello, and Sara Magliacane. “A Sparsity Principle
for Partially Observable Causal Representation Learning.” In Causal Representation
Learning Workshop at NeurIPS 2023. OpenReview, 2023.
ieee: D. Xu et al., “A sparsity principle for partially observable causal
representation learning,” in Causal Representation Learning Workshop at NeurIPS
2023, New Orleans, LA, United States, 2023.
ista: 'Xu D, Yao D, Lachapelle S, Taslakian P, von Kügelgen J, Locatello F, Magliacane
S. 2023. A sparsity principle for partially observable causal representation learning.
Causal Representation Learning Workshop at NeurIPS 2023. CRL: Causal Representation
Learning Workshop at NeurIPS, 54.'
mla: Xu, Danru, et al. “A Sparsity Principle for Partially Observable Causal Representation
Learning.” Causal Representation Learning Workshop at NeurIPS 2023, 54,
OpenReview, 2023.
short: D. Xu, D. Yao, S. Lachapelle, P. Taslakian, J. von Kügelgen, F. Locatello,
S. Magliacane, in:, Causal Representation Learning Workshop at NeurIPS 2023, OpenReview,
2023.
conference:
end_date: 2023-12-15
location: New Orleans, LA, United States
name: 'CRL: Causal Representation Learning Workshop at NeurIPS'
start_date: 2023-12-15
date_created: 2024-02-07T15:17:51Z
date_published: 2023-12-05T00:00:00Z
date_updated: 2024-02-13T08:59:27Z
day: '05'
ddc:
- '000'
department:
- _id: FrLo
file:
- access_level: open_access
checksum: 484efc27bda75ed6666044989695d9b6
content_type: application/pdf
creator: dernst
date_created: 2024-02-13T08:50:53Z
date_updated: 2024-02-13T08:50:53Z
file_id: '14982'
file_name: 2023_CRL_Xu.pdf
file_size: 552357
relation: main_file
success: 1
file_date_updated: 2024-02-13T08:50:53Z
has_accepted_license: '1'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://openreview.net/forum?id=Whr6uobelR
month: '12'
oa: 1
oa_version: Published Version
publication: Causal Representation Learning Workshop at NeurIPS 2023
publication_status: published
publisher: OpenReview
quality_controlled: '1'
status: public
title: A sparsity principle for partially observable causal representation learning
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: '2023'
...
---
_id: '14961'
abstract:
- lang: eng
text: "The use of simulated data in the field of causal discovery is ubiquitous
due to the scarcity of annotated real data. Recently, Reisach et al., 2021 highlighted
the emergence of patterns in simulated linear data, which displays increasing
marginal variance in the casual direction. As an ablation in their experiments,
Montagna et al., 2023 found that similar patterns may emerge in\r\nnonlinear models
for the variance of the score vector $\\nabla \\log p_{\\mathbf{X}}$, and introduced
the ScoreSort algorithm. In this work, we formally define and characterize this
score-sortability pattern of nonlinear additive noise models. We find that it
defines a class of identifiable (bivariate) causal models overlapping with nonlinear
additive noise models. We\r\ntheoretically demonstrate the advantages of ScoreSort
in terms of statistical efficiency compared to prior state-of-the-art score matching-based
methods and empirically show the score-sortability of the most common synthetic
benchmarks in the literature. Our findings remark (1) the lack of diversity in
the data as an important limitation in the evaluation of nonlinear causal discovery
approaches, (2) the importance of thoroughly testing different settings within
a problem class, and (3) the importance of analyzing statistical properties in\r\ncausal
discovery, where research is often limited to defining identifiability conditions
of the model. "
article_number: '2310.14246'
article_processing_charge: No
arxiv: 1
author:
- first_name: Francesco
full_name: Montagna, Francesco
last_name: Montagna
- first_name: Nicoletta
full_name: Noceti, Nicoletta
last_name: Noceti
- first_name: Lorenzo
full_name: Rosasco, Lorenzo
last_name: Rosasco
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: Montagna F, Noceti N, Rosasco L, Locatello F. Shortcuts for causal discovery
of nonlinear models by score matching. arXiv. doi:10.48550/arXiv.2310.14246
apa: Montagna, F., Noceti, N., Rosasco, L., & Locatello, F. (n.d.). Shortcuts
for causal discovery of nonlinear models by score matching. arXiv. https://doi.org/10.48550/arXiv.2310.14246
chicago: Montagna, Francesco, Nicoletta Noceti, Lorenzo Rosasco, and Francesco Locatello.
“Shortcuts for Causal Discovery of Nonlinear Models by Score Matching.” ArXiv,
n.d. https://doi.org/10.48550/arXiv.2310.14246.
ieee: F. Montagna, N. Noceti, L. Rosasco, and F. Locatello, “Shortcuts for causal
discovery of nonlinear models by score matching,” arXiv. .
ista: Montagna F, Noceti N, Rosasco L, Locatello F. Shortcuts for causal discovery
of nonlinear models by score matching. arXiv, 2310.14246.
mla: Montagna, Francesco, et al. “Shortcuts for Causal Discovery of Nonlinear Models
by Score Matching.” ArXiv, 2310.14246, doi:10.48550/arXiv.2310.14246.
short: F. Montagna, N. Noceti, L. Rosasco, F. Locatello, ArXiv (n.d.).
date_created: 2024-02-08T15:31:46Z
date_published: 2023-10-22T00:00:00Z
date_updated: 2024-02-12T10:03:33Z
day: '22'
department:
- _id: FrLo
doi: 10.48550/arXiv.2310.14246
external_id:
arxiv:
- '2310.14246'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2310.14246
month: '10'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Shortcuts for causal discovery of nonlinear models by score matching
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14962'
abstract:
- lang: eng
text: "In this paper, we show that recent advances in video representation learning\r\nand
pre-trained vision-language models allow for substantial improvements in\r\nself-supervised
video object localization. We propose a method that first\r\nlocalizes objects
in videos via a slot attention approach and then assigns text\r\nto the obtained
slots. The latter is achieved by an unsupervised way to read\r\nlocalized semantic
information from the pre-trained CLIP model. The resulting\r\nvideo object localization
is entirely unsupervised apart from the implicit\r\nannotation contained in CLIP,
and it is effectively the first unsupervised\r\napproach that yields good results
on regular video benchmarks."
article_number: '2309.09858'
article_processing_charge: No
arxiv: 1
author:
- first_name: Ke
full_name: Fan, Ke
last_name: Fan
- first_name: Zechen
full_name: Bai, Zechen
last_name: Bai
- first_name: Tianjun
full_name: Xiao, Tianjun
last_name: Xiao
- first_name: Dominik
full_name: Zietlow, Dominik
last_name: Zietlow
- first_name: Max
full_name: Horn, Max
last_name: Horn
- first_name: Zixu
full_name: Zhao, Zixu
last_name: Zhao
- first_name: Carl-Johann Simon-Gabriel
full_name: Carl-Johann Simon-Gabriel, Carl-Johann Simon-Gabriel
last_name: Carl-Johann Simon-Gabriel
- first_name: Mike Zheng
full_name: Shou, Mike Zheng
last_name: Shou
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Bernt
full_name: Schiele, Bernt
last_name: Schiele
- first_name: Thomas
full_name: Brox, Thomas
last_name: Brox
- first_name: Zheng
full_name: Zhang, Zheng
last_name: Zhang
- first_name: Yanwei
full_name: Fu, Yanwei
last_name: Fu
- first_name: Tong
full_name: He, Tong
last_name: He
citation:
ama: Fan K, Bai Z, Xiao T, et al. Unsupervised open-vocabulary object localization
in videos. arXiv. doi:10.48550/arXiv.2309.09858
apa: Fan, K., Bai, Z., Xiao, T., Zietlow, D., Horn, M., Zhao, Z., … He, T. (n.d.).
Unsupervised open-vocabulary object localization in videos. arXiv. https://doi.org/10.48550/arXiv.2309.09858
chicago: Fan, Ke, Zechen Bai, Tianjun Xiao, Dominik Zietlow, Max Horn, Zixu Zhao,
Carl-Johann Simon-Gabriel Carl-Johann Simon-Gabriel, et al. “Unsupervised Open-Vocabulary
Object Localization in Videos.” ArXiv, n.d. https://doi.org/10.48550/arXiv.2309.09858.
ieee: K. Fan et al., “Unsupervised open-vocabulary object localization in
videos,” arXiv. .
ista: Fan K, Bai Z, Xiao T, Zietlow D, Horn M, Zhao Z, Carl-Johann Simon-Gabriel
C-JS-G, Shou MZ, Locatello F, Schiele B, Brox T, Zhang Z, Fu Y, He T. Unsupervised
open-vocabulary object localization in videos. arXiv, 2309.09858.
mla: Fan, Ke, et al. “Unsupervised Open-Vocabulary Object Localization in Videos.”
ArXiv, 2309.09858, doi:10.48550/arXiv.2309.09858.
short: K. Fan, Z. Bai, T. Xiao, D. Zietlow, M. Horn, Z. Zhao, C.-J.S.-G. Carl-Johann
Simon-Gabriel, M.Z. Shou, F. Locatello, B. Schiele, T. Brox, Z. Zhang, Y. Fu,
T. He, ArXiv (n.d.).
date_created: 2024-02-08T15:33:39Z
date_published: 2023-09-18T00:00:00Z
date_updated: 2024-02-12T10:12:22Z
day: '18'
department:
- _id: FrLo
doi: 10.48550/arXiv.2309.09858
extern: '1'
external_id:
arxiv:
- '2309.09858'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2309.09858
month: '09'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Unsupervised open-vocabulary object localization in videos
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14963'
abstract:
- lang: eng
text: "Unsupervised object-centric learning methods allow the partitioning of scenes\r\ninto
entities without additional localization information and are excellent\r\ncandidates
for reducing the annotation burden of multiple-object tracking (MOT)\r\npipelines.
Unfortunately, they lack two key properties: objects are often split\r\ninto parts
and are not consistently tracked over time. In fact,\r\nstate-of-the-art models
achieve pixel-level accuracy and temporal consistency\r\nby relying on supervised
object detection with additional ID labels for the\r\nassociation through time.
This paper proposes a video object-centric model for\r\nMOT. It consists of an
index-merge module that adapts the object-centric slots\r\ninto detection outputs
and an object memory module that builds complete object\r\nprototypes to handle
occlusions. Benefited from object-centric learning, we\r\nonly require sparse
detection labels (0%-6.25%) for object localization and\r\nfeature binding. Relying
on our self-supervised\r\nExpectation-Maximization-inspired loss for object association,
our approach\r\nrequires no ID labels. Our experiments significantly narrow the
gap between the\r\nexisting object-centric model and the fully supervised state-of-the-art
and\r\noutperform several unsupervised trackers."
article_number: '2309.00233'
article_processing_charge: No
arxiv: 1
author:
- first_name: Zixu
full_name: Zhao, Zixu
last_name: Zhao
- first_name: Jiaze
full_name: Wang, Jiaze
last_name: Wang
- first_name: Max
full_name: Horn, Max
last_name: Horn
- first_name: Yizhuo
full_name: Ding, Yizhuo
last_name: Ding
- first_name: Tong
full_name: He, Tong
last_name: He
- first_name: Zechen
full_name: Bai, Zechen
last_name: Bai
- first_name: Dominik
full_name: Zietlow, Dominik
last_name: Zietlow
- first_name: Carl-Johann Simon-Gabriel
full_name: Carl-Johann Simon-Gabriel, Carl-Johann Simon-Gabriel
last_name: Carl-Johann Simon-Gabriel
- first_name: Bing
full_name: Shuai, Bing
last_name: Shuai
- first_name: Zhuowen
full_name: Tu, Zhuowen
last_name: Tu
- first_name: Thomas
full_name: Brox, Thomas
last_name: Brox
- first_name: Bernt
full_name: Schiele, Bernt
last_name: Schiele
- first_name: Yanwei
full_name: Fu, Yanwei
last_name: Fu
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Zheng
full_name: Zhang, Zheng
last_name: Zhang
- first_name: Tianjun
full_name: Xiao, Tianjun
last_name: Xiao
citation:
ama: Zhao Z, Wang J, Horn M, et al. Object-centric multiple object tracking. arXiv.
doi:10.48550/arXiv.2309.00233
apa: Zhao, Z., Wang, J., Horn, M., Ding, Y., He, T., Bai, Z., … Xiao, T. (n.d.).
Object-centric multiple object tracking. arXiv. https://doi.org/10.48550/arXiv.2309.00233
chicago: Zhao, Zixu, Jiaze Wang, Max Horn, Yizhuo Ding, Tong He, Zechen Bai, Dominik
Zietlow, et al. “Object-Centric Multiple Object Tracking.” ArXiv, n.d.
https://doi.org/10.48550/arXiv.2309.00233.
ieee: Z. Zhao et al., “Object-centric multiple object tracking,” arXiv.
.
ista: Zhao Z, Wang J, Horn M, Ding Y, He T, Bai Z, Zietlow D, Carl-Johann Simon-Gabriel
C-JS-G, Shuai B, Tu Z, Brox T, Schiele B, Fu Y, Locatello F, Zhang Z, Xiao T.
Object-centric multiple object tracking. arXiv, 2309.00233.
mla: Zhao, Zixu, et al. “Object-Centric Multiple Object Tracking.” ArXiv,
2309.00233, doi:10.48550/arXiv.2309.00233.
short: Z. Zhao, J. Wang, M. Horn, Y. Ding, T. He, Z. Bai, D. Zietlow, C.-J.S.-G.
Carl-Johann Simon-Gabriel, B. Shuai, Z. Tu, T. Brox, B. Schiele, Y. Fu, F. Locatello,
Z. Zhang, T. Xiao, ArXiv (n.d.).
date_created: 2024-02-08T15:34:43Z
date_published: 2023-09-01T00:00:00Z
date_updated: 2024-02-12T10:16:21Z
day: '01'
department:
- _id: FrLo
doi: 10.48550/arXiv.2309.00233
extern: '1'
external_id:
arxiv:
- '2309.00233'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: ' https://doi.org/10.48550/arXiv.2309.00233'
month: '09'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Object-centric multiple object tracking
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14105'
abstract:
- lang: eng
text: "Despite their recent success, deep neural networks continue to perform poorly
when they encounter distribution shifts at test time. Many recently proposed approaches
try to counter this by aligning the model to the new distribution prior to inference.
With no labels available this requires unsupervised objectives to adapt the model
on the observed test data. In this paper, we propose Test-Time SelfTraining (TeST):
a technique that takes as input a model trained on some source data and a novel
data distribution at test time, and learns invariant and robust representations
using a student-teacher framework. We find that models adapted using TeST significantly
improve over baseline testtime adaptation algorithms. TeST achieves competitive
performance to modern domain adaptation algorithms [4, 43], while having access
to 5-10x less data at time of adaption. We thoroughly evaluate a variety of baselines
on two tasks:\r\nobject detection and image segmentation and find that models
adapted with TeST. We find that TeST sets the new stateof-the art for test-time
domain adaptation algorithms. "
article_processing_charge: No
arxiv: 1
author:
- first_name: Samarth
full_name: Sinha, Samarth
last_name: Sinha
- first_name: Peter
full_name: Gehler, Peter
last_name: Gehler
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Bernt
full_name: Schiele, Bernt
last_name: Schiele
citation:
ama: 'Sinha S, Gehler P, Locatello F, Schiele B. TeST: Test-time Self-Training under
distribution shift. In: 2023 IEEE/CVF Winter Conference on Applications of
Computer Vision. Institute of Electrical and Electronics Engineers; 2023.
doi:10.1109/wacv56688.2023.00278'
apa: 'Sinha, S., Gehler, P., Locatello, F., & Schiele, B. (2023). TeST: Test-time
Self-Training under distribution shift. In 2023 IEEE/CVF Winter Conference
on Applications of Computer Vision. Waikoloa, HI, United States: Institute
of Electrical and Electronics Engineers. https://doi.org/10.1109/wacv56688.2023.00278'
chicago: 'Sinha, Samarth, Peter Gehler, Francesco Locatello, and Bernt Schiele.
“TeST: Test-Time Self-Training under Distribution Shift.” In 2023 IEEE/CVF
Winter Conference on Applications of Computer Vision. Institute of Electrical
and Electronics Engineers, 2023. https://doi.org/10.1109/wacv56688.2023.00278.'
ieee: 'S. Sinha, P. Gehler, F. Locatello, and B. Schiele, “TeST: Test-time Self-Training
under distribution shift,” in 2023 IEEE/CVF Winter Conference on Applications
of Computer Vision, Waikoloa, HI, United States, 2023.'
ista: 'Sinha S, Gehler P, Locatello F, Schiele B. 2023. TeST: Test-time Self-Training
under distribution shift. 2023 IEEE/CVF Winter Conference on Applications of Computer
Vision. WACV: Winter Conference on Applications of Computer Vision.'
mla: 'Sinha, Samarth, et al. “TeST: Test-Time Self-Training under Distribution Shift.”
2023 IEEE/CVF Winter Conference on Applications of Computer Vision, Institute
of Electrical and Electronics Engineers, 2023, doi:10.1109/wacv56688.2023.00278.'
short: S. Sinha, P. Gehler, F. Locatello, B. Schiele, in:, 2023 IEEE/CVF Winter
Conference on Applications of Computer Vision, Institute of Electrical and Electronics
Engineers, 2023.
conference:
end_date: 2023-01-07
location: Waikoloa, HI, United States
name: 'WACV: Winter Conference on Applications of Computer Vision'
start_date: 2023-01-02
date_created: 2023-08-21T12:11:38Z
date_published: 2023-02-06T00:00:00Z
date_updated: 2023-09-06T10:26:56Z
day: '06'
department:
- _id: FrLo
doi: 10.1109/wacv56688.2023.00278
extern: '1'
external_id:
arxiv:
- '2209.11459'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2209.11459
month: '02'
oa: 1
oa_version: Preprint
publication: 2023 IEEE/CVF Winter Conference on Applications of Computer Vision
publication_identifier:
eissn:
- 2642-9381
isbn:
- '9781665493475'
publication_status: published
publisher: Institute of Electrical and Electronics Engineers
quality_controlled: '1'
scopus_import: '1'
status: public
title: 'TeST: Test-time Self-Training under distribution shift'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14207'
abstract:
- lang: eng
text: The binding problem in human cognition, concerning how the brain represents
and connects objects within a fixed network of neural connections, remains a subject
of intense debate. Most machine learning efforts addressing this issue in an unsupervised
setting have focused on slot-based methods, which may be limiting due to their
discrete nature and difficulty to express uncertainty. Recently, the Complex AutoEncoder
was proposed as an alternative that learns continuous and distributed object-centric
representations. However, it is only applicable to simple toy data. In this paper,
we present Rotating Features, a generalization of complex-valued features to higher
dimensions, and a new evaluation procedure for extracting objects from distributed
representations. Additionally, we show the applicability of our approach to pre-trained
features. Together, these advancements enable us to scale distributed object-centric
representations from simple toy to real-world data. We believe this work advances
a new paradigm for addressing the binding problem in machine learning and has
the potential to inspire further innovation in the field.
article_number: '2306.00600'
article_processing_charge: No
arxiv: 1
author:
- first_name: Sindy
full_name: Löwe, Sindy
last_name: Löwe
- first_name: Phillip
full_name: Lippe, Phillip
last_name: Lippe
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Max
full_name: Welling, Max
last_name: Welling
citation:
ama: Löwe S, Lippe P, Locatello F, Welling M. Rotating features for object discovery.
arXiv. doi:10.48550/arXiv.2306.00600
apa: Löwe, S., Lippe, P., Locatello, F., & Welling, M. (n.d.). Rotating features
for object discovery. arXiv. https://doi.org/10.48550/arXiv.2306.00600
chicago: Löwe, Sindy, Phillip Lippe, Francesco Locatello, and Max Welling. “Rotating
Features for Object Discovery.” ArXiv, n.d. https://doi.org/10.48550/arXiv.2306.00600.
ieee: S. Löwe, P. Lippe, F. Locatello, and M. Welling, “Rotating features for object
discovery,” arXiv. .
ista: Löwe S, Lippe P, Locatello F, Welling M. Rotating features for object discovery.
arXiv, 2306.00600.
mla: Löwe, Sindy, et al. “Rotating Features for Object Discovery.” ArXiv,
2306.00600, doi:10.48550/arXiv.2306.00600.
short: S. Löwe, P. Lippe, F. Locatello, M. Welling, ArXiv (n.d.).
date_created: 2023-08-22T14:18:00Z
date_published: 2023-06-01T00:00:00Z
date_updated: 2024-02-12T09:53:44Z
day: '01'
department:
- _id: FrLo
doi: 10.48550/arXiv.2306.00600
external_id:
arxiv:
- '2306.00600'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2306.00600
month: '06'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Rotating features for object discovery
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14208'
abstract:
- lang: eng
text: This paper focuses on over-parameterized deep neural networks (DNNs) with
ReLU activation functions and proves that when the data distribution is well-separated,
DNNs can achieve Bayes-optimal test error for classification while obtaining (nearly)
zero-training error under the lazy training regime. For this purpose, we unify
three interrelated concepts of overparameterization, benign overfitting, and the
Lipschitz constant of DNNs. Our results indicate that interpolating with smoother
functions leads to better generalization. Furthermore, we investigate the special
case where interpolating smooth ground-truth functions is performed by DNNs under
the Neural Tangent Kernel (NTK) regime for generalization. Our result demonstrates
that the generalization error converges to a constant order that only depends
on label noise and initialization noise, which theoretically verifies benign overfitting.
Our analysis provides a tight lower bound on the normalized margin under non-smooth
activation functions, as well as the minimum eigenvalue of NTK under high-dimensional
settings, which has its own interest in learning theory.
alternative_title:
- PMLR
article_processing_charge: No
arxiv: 1
author:
- first_name: Zhenyu
full_name: Zhu, Zhenyu
last_name: Zhu
- first_name: Fanghui
full_name: Liu, Fanghui
last_name: Liu
- first_name: Grigorios G
full_name: Chrysos, Grigorios G
last_name: Chrysos
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Volkan
full_name: Cevher, Volkan
last_name: Cevher
citation:
ama: 'Zhu Z, Liu F, Chrysos GG, Locatello F, Cevher V. Benign overfitting in deep
neural networks under lazy training. In: Proceedings of the 40th International
Conference on Machine Learning. Vol 202. ML Research Press; 2023:43105-43128.'
apa: 'Zhu, Z., Liu, F., Chrysos, G. G., Locatello, F., & Cevher, V. (2023).
Benign overfitting in deep neural networks under lazy training. In Proceedings
of the 40th International Conference on Machine Learning (Vol. 202, pp. 43105–43128).
Honolulu, Hawaii, United States: ML Research Press.'
chicago: Zhu, Zhenyu, Fanghui Liu, Grigorios G Chrysos, Francesco Locatello, and
Volkan Cevher. “Benign Overfitting in Deep Neural Networks under Lazy Training.”
In Proceedings of the 40th International Conference on Machine Learning,
202:43105–28. ML Research Press, 2023.
ieee: Z. Zhu, F. Liu, G. G. Chrysos, F. Locatello, and V. Cevher, “Benign overfitting
in deep neural networks under lazy training,” in Proceedings of the 40th International
Conference on Machine Learning, Honolulu, Hawaii, United States, 2023, vol.
202, pp. 43105–43128.
ista: Zhu Z, Liu F, Chrysos GG, Locatello F, Cevher V. 2023. Benign overfitting
in deep neural networks under lazy training. Proceedings of the 40th International
Conference on Machine Learning. International Conference on Machine Learning,
PMLR, vol. 202, 43105–43128.
mla: Zhu, Zhenyu, et al. “Benign Overfitting in Deep Neural Networks under Lazy
Training.” Proceedings of the 40th International Conference on Machine Learning,
vol. 202, ML Research Press, 2023, pp. 43105–28.
short: Z. Zhu, F. Liu, G.G. Chrysos, F. Locatello, V. Cevher, in:, Proceedings of
the 40th International Conference on Machine Learning, ML Research Press, 2023,
pp. 43105–43128.
conference:
end_date: 2023-07-29
location: Honolulu, Hawaii, United States
name: International Conference on Machine Learning
start_date: 2023-07-23
date_created: 2023-08-22T14:18:18Z
date_published: 2023-05-30T00:00:00Z
date_updated: 2023-09-13T08:46:46Z
day: '30'
department:
- _id: FrLo
extern: '1'
external_id:
arxiv:
- '2305.19377'
intvolume: ' 202'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2305.19377
month: '05'
oa: 1
oa_version: Preprint
page: 43105-43128
publication: Proceedings of the 40th International Conference on Machine Learning
publication_status: published
publisher: ML Research Press
quality_controlled: '1'
status: public
title: Benign overfitting in deep neural networks under lazy training
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 202
year: '2023'
...
---
_id: '14209'
abstract:
- lang: eng
text: Diffusion models excel at generating photorealistic images from text-queries.
Naturally, many approaches have been proposed to use these generative abilities
to augment training datasets for downstream tasks, such as classification. However,
diffusion models are themselves trained on large noisily supervised, but nonetheless,
annotated datasets. It is an open question whether the generalization capabilities
of diffusion models beyond using the additional data of the pre-training process
for augmentation lead to improved downstream performance. We perform a systematic
evaluation of existing methods to generate images from diffusion models and study
new extensions to assess their benefit for data augmentation. While we find that
personalizing diffusion models towards the target data outperforms simpler prompting
strategies, we also show that using the training data of the diffusion model alone,
via a simple nearest neighbor retrieval procedure, leads to even stronger downstream
performance. Overall, our study probes the limitations of diffusion models for
data augmentation but also highlights its potential in generating new training
data to improve performance on simple downstream vision tasks.
article_number: '2304.10253'
article_processing_charge: No
arxiv: 1
author:
- first_name: Max F.
full_name: Burg, Max F.
last_name: Burg
- first_name: Florian
full_name: Wenzel, Florian
last_name: Wenzel
- first_name: Dominik
full_name: Zietlow, Dominik
last_name: Zietlow
- first_name: Max
full_name: Horn, Max
last_name: Horn
- first_name: Osama
full_name: Makansi, Osama
last_name: Makansi
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
- first_name: Chris
full_name: Russell, Chris
last_name: Russell
citation:
ama: Burg MF, Wenzel F, Zietlow D, et al. A data augmentation perspective on diffusion
models and retrieval. arXiv. doi:10.48550/arXiv.2304.10253
apa: Burg, M. F., Wenzel, F., Zietlow, D., Horn, M., Makansi, O., Locatello, F.,
& Russell, C. (n.d.). A data augmentation perspective on diffusion models
and retrieval. arXiv. https://doi.org/10.48550/arXiv.2304.10253
chicago: Burg, Max F., Florian Wenzel, Dominik Zietlow, Max Horn, Osama Makansi,
Francesco Locatello, and Chris Russell. “A Data Augmentation Perspective on Diffusion
Models and Retrieval.” ArXiv, n.d. https://doi.org/10.48550/arXiv.2304.10253.
ieee: M. F. Burg et al., “A data augmentation perspective on diffusion models
and retrieval,” arXiv. .
ista: Burg MF, Wenzel F, Zietlow D, Horn M, Makansi O, Locatello F, Russell C. A
data augmentation perspective on diffusion models and retrieval. arXiv, 2304.10253.
mla: Burg, Max F., et al. “A Data Augmentation Perspective on Diffusion Models and
Retrieval.” ArXiv, 2304.10253, doi:10.48550/arXiv.2304.10253.
short: M.F. Burg, F. Wenzel, D. Zietlow, M. Horn, O. Makansi, F. Locatello, C. Russell,
ArXiv (n.d.).
date_created: 2023-08-22T14:18:43Z
date_published: 2023-04-20T00:00:00Z
date_updated: 2023-09-13T08:51:56Z
day: '20'
department:
- _id: FrLo
doi: 10.48550/arXiv.2304.10253
extern: '1'
external_id:
arxiv:
- '2304.10253'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2304.10253
month: '04'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: A data augmentation perspective on diffusion models and retrieval
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14210'
abstract:
- lang: eng
text: Recovering the latent factors of variation of high dimensional data has so
far focused on simple synthetic settings. Mostly building on unsupervised and
weakly-supervised objectives, prior work missed out on the positive implications
for representation learning on real world data. In this work, we propose to leverage
knowledge extracted from a diversified set of supervised tasks to learn a common
disentangled representation. Assuming each supervised task only depends on an
unknown subset of the factors of variation, we disentangle the feature space of
a supervised multi-task model, with features activating sparsely across different
tasks and information being shared as appropriate. Importantly, we never directly
observe the factors of variations but establish that access to multiple tasks
is sufficient for identifiability under sufficiency and minimality assumptions.
We validate our approach on six real world distribution shift benchmarks, and
different data modalities (images, text), demonstrating how disentangled representations
can be transferred to real settings.
article_number: '2304.07939'
article_processing_charge: No
arxiv: 1
author:
- first_name: Marco
full_name: Fumero, Marco
last_name: Fumero
- first_name: Florian
full_name: Wenzel, Florian
last_name: Wenzel
- first_name: Luca
full_name: Zancato, Luca
last_name: Zancato
- first_name: Alessandro
full_name: Achille, Alessandro
last_name: Achille
- first_name: Emanuele
full_name: Rodolà, Emanuele
last_name: Rodolà
- first_name: Stefano
full_name: Soatto, Stefano
last_name: Soatto
- first_name: Bernhard
full_name: Schölkopf, Bernhard
last_name: Schölkopf
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: Fumero M, Wenzel F, Zancato L, et al. Leveraging sparse and shared feature
activations for disentangled representation learning. arXiv. doi:10.48550/arXiv.2304.07939
apa: Fumero, M., Wenzel, F., Zancato, L., Achille, A., Rodolà, E., Soatto, S., …
Locatello, F. (n.d.). Leveraging sparse and shared feature activations for disentangled
representation learning. arXiv. https://doi.org/10.48550/arXiv.2304.07939
chicago: Fumero, Marco, Florian Wenzel, Luca Zancato, Alessandro Achille, Emanuele
Rodolà, Stefano Soatto, Bernhard Schölkopf, and Francesco Locatello. “Leveraging
Sparse and Shared Feature Activations for Disentangled Representation Learning.”
ArXiv, n.d. https://doi.org/10.48550/arXiv.2304.07939.
ieee: M. Fumero et al., “Leveraging sparse and shared feature activations
for disentangled representation learning,” arXiv. .
ista: Fumero M, Wenzel F, Zancato L, Achille A, Rodolà E, Soatto S, Schölkopf B,
Locatello F. Leveraging sparse and shared feature activations for disentangled
representation learning. arXiv, 2304.07939.
mla: Fumero, Marco, et al. “Leveraging Sparse and Shared Feature Activations for
Disentangled Representation Learning.” ArXiv, 2304.07939, doi:10.48550/arXiv.2304.07939.
short: M. Fumero, F. Wenzel, L. Zancato, A. Achille, E. Rodolà, S. Soatto, B. Schölkopf,
F. Locatello, ArXiv (n.d.).
date_created: 2023-08-22T14:19:03Z
date_published: 2023-04-17T00:00:00Z
date_updated: 2024-02-12T09:55:48Z
day: '17'
department:
- _id: FrLo
doi: 10.48550/arXiv.2304.07939
external_id:
arxiv:
- '2304.07939'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2304.07939
month: '04'
oa: 1
oa_version: Preprint
publication: arXiv
publication_status: submitted
status: public
title: Leveraging sparse and shared feature activations for disentangled representation
learning
type: preprint
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14211'
abstract:
- lang: eng
text: 'Causal discovery methods are intrinsically constrained by the set of assumptions
needed to ensure structure identifiability. Moreover additional restrictions are
often imposed in order to simplify the inference task: this is the case for the
Gaussian noise assumption on additive non-linear models, which is common to many
causal discovery approaches. In this paper we show the shortcomings of inference
under this hypothesis, analyzing the risk of edge inversion under violation of
Gaussianity of the noise terms. Then, we propose a novel method for inferring
the topological ordering of the variables in the causal graph, from data generated
according to an additive non-linear model with a generic noise distribution. This
leads to NoGAM (Not only Gaussian Additive noise Models), a causal discovery algorithm
with a minimal set of assumptions and state of the art performance, experimentally
benchmarked on synthetic data.'
article_processing_charge: No
arxiv: 1
author:
- first_name: Francesco
full_name: Montagna, Francesco
last_name: Montagna
- first_name: Nicoletta
full_name: Noceti, Nicoletta
last_name: Noceti
- first_name: Lorenzo
full_name: Rosasco, Lorenzo
last_name: Rosasco
- first_name: Kun
full_name: Zhang, Kun
last_name: Zhang
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: 'Montagna F, Noceti N, Rosasco L, Zhang K, Locatello F. Causal discovery with
score matching on additive models with arbitrary noise. In: 2nd Conference
on Causal Learning and Reasoning. ; 2023.'
apa: Montagna, F., Noceti, N., Rosasco, L., Zhang, K., & Locatello, F. (2023).
Causal discovery with score matching on additive models with arbitrary noise.
In 2nd Conference on Causal Learning and Reasoning. Tübingen, Germany.
chicago: Montagna, Francesco, Nicoletta Noceti, Lorenzo Rosasco, Kun Zhang, and
Francesco Locatello. “Causal Discovery with Score Matching on Additive Models
with Arbitrary Noise.” In 2nd Conference on Causal Learning and Reasoning,
2023.
ieee: F. Montagna, N. Noceti, L. Rosasco, K. Zhang, and F. Locatello, “Causal discovery
with score matching on additive models with arbitrary noise,” in 2nd Conference
on Causal Learning and Reasoning, Tübingen, Germany, 2023.
ista: 'Montagna F, Noceti N, Rosasco L, Zhang K, Locatello F. 2023. Causal discovery
with score matching on additive models with arbitrary noise. 2nd Conference on
Causal Learning and Reasoning. CLeaR: Conference on Causal Learning and Reasoning.'
mla: Montagna, Francesco, et al. “Causal Discovery with Score Matching on Additive
Models with Arbitrary Noise.” 2nd Conference on Causal Learning and Reasoning,
2023.
short: F. Montagna, N. Noceti, L. Rosasco, K. Zhang, F. Locatello, in:, 2nd Conference
on Causal Learning and Reasoning, 2023.
conference:
end_date: 2023-04-14
location: Tübingen, Germany
name: 'CLeaR: Conference on Causal Learning and Reasoning'
start_date: 2023-04-11
date_created: 2023-08-22T14:19:21Z
date_published: 2023-04-01T00:00:00Z
date_updated: 2023-09-13T09:00:31Z
day: '01'
department:
- _id: FrLo
extern: '1'
external_id:
arxiv:
- '2304.03265'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2304.03265
month: '04'
oa: 1
oa_version: Preprint
publication: 2nd Conference on Causal Learning and Reasoning
publication_status: published
quality_controlled: '1'
scopus_import: '1'
status: public
title: Causal discovery with score matching on additive models with arbitrary noise
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14212'
abstract:
- lang: eng
text: This paper demonstrates how to discover the whole causal graph from the second
derivative of the log-likelihood in observational non-linear additive Gaussian
noise models. Leveraging scalable machine learning approaches to approximate the
score function ∇logp(X), we extend the work of Rolland et al. (2022) that only
recovers the topological order from the score and requires an expensive pruning
step removing spurious edges among those admitted by the ordering. Our analysis
leads to DAS (acronym for Discovery At Scale), a practical algorithm that reduces
the complexity of the pruning by a factor proportional to the graph size. In practice,
DAS achieves competitive accuracy with current state-of-the-art while being over
an order of magnitude faster. Overall, our approach enables principled and scalable
causal discovery, significantly lowering the compute bar.
article_processing_charge: No
arxiv: 1
author:
- first_name: Francesco
full_name: Montagna, Francesco
last_name: Montagna
- first_name: Nicoletta
full_name: Noceti, Nicoletta
last_name: Noceti
- first_name: Lorenzo
full_name: Rosasco, Lorenzo
last_name: Rosasco
- first_name: Kun
full_name: Zhang, Kun
last_name: Zhang
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: 'Montagna F, Noceti N, Rosasco L, Zhang K, Locatello F. Scalable causal discovery
with score matching. In: 2nd Conference on Causal Learning and Reasoning.
; 2023.'
apa: Montagna, F., Noceti, N., Rosasco, L., Zhang, K., & Locatello, F. (2023).
Scalable causal discovery with score matching. In 2nd Conference on Causal
Learning and Reasoning. Tübingen, Germany.
chicago: Montagna, Francesco, Nicoletta Noceti, Lorenzo Rosasco, Kun Zhang, and
Francesco Locatello. “Scalable Causal Discovery with Score Matching.” In 2nd
Conference on Causal Learning and Reasoning, 2023.
ieee: F. Montagna, N. Noceti, L. Rosasco, K. Zhang, and F. Locatello, “Scalable
causal discovery with score matching,” in 2nd Conference on Causal Learning
and Reasoning, Tübingen, Germany, 2023.
ista: 'Montagna F, Noceti N, Rosasco L, Zhang K, Locatello F. 2023. Scalable causal
discovery with score matching. 2nd Conference on Causal Learning and Reasoning.
CLeaR: Conference on Causal Learning and Reasoning.'
mla: Montagna, Francesco, et al. “Scalable Causal Discovery with Score Matching.”
2nd Conference on Causal Learning and Reasoning, 2023.
short: F. Montagna, N. Noceti, L. Rosasco, K. Zhang, F. Locatello, in:, 2nd Conference
on Causal Learning and Reasoning, 2023.
conference:
end_date: 2023-04-14
location: Tübingen, Germany
name: 'CLeaR: Conference on Causal Learning and Reasoning'
start_date: 2023-04-11
date_created: 2023-08-22T14:19:40Z
date_published: 2023-04-01T00:00:00Z
date_updated: 2023-09-13T09:03:24Z
day: '01'
department:
- _id: FrLo
extern: '1'
external_id:
arxiv:
- '2304.03382'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://arxiv.org/abs/2304.03382
month: '04'
oa: 1
oa_version: Preprint
publication: 2nd Conference on Causal Learning and Reasoning
publication_status: published
quality_controlled: '1'
scopus_import: '1'
status: public
title: Scalable causal discovery with score matching
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...
---
_id: '14214'
abstract:
- lang: eng
text: 'Recent years have seen a surge of interest in learning high-level causal
representations from low-level image pairs under interventions. Yet, existing
efforts are largely limited to simple synthetic settings that are far away from
real-world problems. In this paper, we present Causal Triplet, a causal representation
learning benchmark featuring not only visually more complex scenes, but also two
crucial desiderata commonly overlooked in previous works: (i) an actionable counterfactual
setting, where only certain object-level variables allow for counterfactual observations
whereas others do not; (ii) an interventional downstream task with an emphasis
on out-of-distribution robustness from the independent causal mechanisms principle.
Through extensive experiments, we find that models built with the knowledge of
disentangled or object-centric representations significantly outperform their
distributed counterparts. However, recent causal representation learning methods
still struggle to identify such latent structures, indicating substantial challenges
and opportunities for future work.'
article_processing_charge: No
arxiv: 1
author:
- first_name: Yuejiang
full_name: Liu, Yuejiang
last_name: Liu
- first_name: Alexandre
full_name: Alahi, Alexandre
last_name: Alahi
- first_name: Chris
full_name: Russell, Chris
last_name: Russell
- first_name: Max
full_name: Horn, Max
last_name: Horn
- first_name: Dominik
full_name: Zietlow, Dominik
last_name: Zietlow
- first_name: Bernhard
full_name: Schölkopf, Bernhard
last_name: Schölkopf
- first_name: Francesco
full_name: Locatello, Francesco
id: 26cfd52f-2483-11ee-8040-88983bcc06d4
last_name: Locatello
orcid: 0000-0002-4850-0683
citation:
ama: 'Liu Y, Alahi A, Russell C, et al. Causal triplet: An open challenge for intervention-centric
causal representation learning. In: 2nd Conference on Causal Learning and Reasoning.
; 2023.'
apa: 'Liu, Y., Alahi, A., Russell, C., Horn, M., Zietlow, D., Schölkopf, B., &
Locatello, F. (2023). Causal triplet: An open challenge for intervention-centric
causal representation learning. In 2nd Conference on Causal Learning and Reasoning.
Tübingen, Germany.'
chicago: 'Liu, Yuejiang, Alexandre Alahi, Chris Russell, Max Horn, Dominik Zietlow,
Bernhard Schölkopf, and Francesco Locatello. “Causal Triplet: An Open Challenge
for Intervention-Centric Causal Representation Learning.” In 2nd Conference
on Causal Learning and Reasoning, 2023.'
ieee: 'Y. Liu et al., “Causal triplet: An open challenge for intervention-centric
causal representation learning,” in 2nd Conference on Causal Learning and Reasoning,
Tübingen, Germany, 2023.'
ista: 'Liu Y, Alahi A, Russell C, Horn M, Zietlow D, Schölkopf B, Locatello F. 2023.
Causal triplet: An open challenge for intervention-centric causal representation
learning. 2nd Conference on Causal Learning and Reasoning. CLeaR: Conference on
Causal Learning and Reasoning.'
mla: 'Liu, Yuejiang, et al. “Causal Triplet: An Open Challenge for Intervention-Centric
Causal Representation Learning.” 2nd Conference on Causal Learning and Reasoning,
2023.'
short: Y. Liu, A. Alahi, C. Russell, M. Horn, D. Zietlow, B. Schölkopf, F. Locatello,
in:, 2nd Conference on Causal Learning and Reasoning, 2023.
conference:
end_date: 2023-04-14
location: Tübingen, Germany
name: 'CLeaR: Conference on Causal Learning and Reasoning'
start_date: 2023-04-11
date_created: 2023-08-22T14:20:18Z
date_published: 2023-04-12T00:00:00Z
date_updated: 2023-09-13T09:23:08Z
day: '12'
department:
- _id: FrLo
extern: '1'
external_id:
arxiv:
- '2301.05169'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://doi.org/10.48550/arXiv.2301.05169
month: '04'
oa: 1
oa_version: Preprint
publication: 2nd Conference on Causal Learning and Reasoning
publication_status: published
quality_controlled: '1'
status: public
title: 'Causal triplet: An open challenge for intervention-centric causal representation
learning'
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
year: '2023'
...