@article{13352,
  abstract     = {Optoelectronic effects differentiating absorption of right and left circularly polarized photons in thin films of chiral materials are typically prohibitively small for their direct photocurrent observation. Chiral metasurfaces increase the electronic sensitivity to circular polarization, but their out-of-plane architecture entails manufacturing and performance trade-offs. Here, we show that nanoporous thin films of chiral nanoparticles enable high sensitivity to circular polarization due to light-induced polarization-dependent ion accumulation at nanoparticle interfaces. Self-assembled multilayers of gold nanoparticles modified with L-phenylalanine generate a photocurrent under right-handed circularly polarized light as high as 2.41 times higher than under left-handed circularly polarized light. The strong plasmonic coupling between the multiple nanoparticles producing planar chiroplasmonic modes facilitates the ejection of electrons, whose entrapment at the membrane–electrolyte interface is promoted by a thick layer of enantiopure phenylalanine. Demonstrated detection of light ellipticity with equal sensitivity at all incident angles mimics phenomenological aspects of polarization vision in marine animals. The simplicity of self-assembly and sensitivity of polarization detection found in optoionic membranes opens the door to a family of miniaturized fluidic devices for chiral photonics.},
  author       = {Cai, Jiarong and Zhang, Wei and Xu, Liguang and Hao, Changlong and Ma, Wei and Sun, Maozhong and Wu, Xiaoling and Qin, Xian and Colombari, Felippe Mariano and de Moura, André Farias and Xu, Jiahui and Silva, Mariana Cristina and Carneiro-Neto, Evaldo Batista and Gomes, Weverson Rodrigues and Vallée, Renaud A. L. and Pereira, Ernesto Chaves and Liu, Xiaogang and Xu, Chuanlai and Klajn, Rafal and Kotov, Nicholas A. and Kuang, Hua},
  issn         = {1748-3395},
  journal      = {Nature Nanotechnology},
  keywords     = {Electrical and Electronic Engineering, Condensed Matter Physics, General Materials Science, Biomedical Engineering, Atomic and Molecular Physics, and Optics, Bioengineering},
  number       = {4},
  pages        = {408--416},
  publisher    = {Springer Nature},
  title        = {{Polarization-sensitive optoionic membranes from chiral plasmonic nanoparticles}},
  doi          = {10.1038/s41565-022-01079-3},
  volume       = {17},
  year         = {2022},
}

@article{13353,
  abstract     = {We show that the optical properties of indigo carmine can be modulated by encapsulation within a coordination cage. Depending on the host/guest molar ratio, the cage can predominantly encapsulate either one or two dye molecules. The 1 : 1 complex is fluorescent, unique for an indigo dye in an aqueous solution. We have also found that binding two dye molecules stabilizes a previously unknown conformation of the cage.},
  author       = {Yanshyna, Oksana and Avram, Liat and Shimon, Linda J. W. and Klajn, Rafal},
  issn         = {1364-548X},
  journal      = {Chemical Communications},
  keywords     = {Materials Chemistry, Metals and Alloys, Surfaces, Coatings and Films, General Chemistry, Ceramics and Composites, Electronic, Optical and Magnetic Materials, Catalysis},
  number       = {21},
  pages        = {3461--3464},
  publisher    = {Royal Society of Chemistry},
  title        = {{Coexistence of 1:1 and 2:1 inclusion complexes of indigo carmine}},
  doi          = {10.1039/d1cc07081a},
  volume       = {58},
  year         = {2022},
}

@article{13355,
  abstract     = {Supramolecular self-assembly in biological systems holds promise to convert and amplify disease-specific signals to physical or mechanical signals that can direct cell fate. However, it remains challenging to design physiologically stable self-assembling systems that demonstrate tunable and predictable behavior. Here, the use of zwitterionic tetrapeptide modalities to direct nanoparticle assembly under physiological conditions is reported. The self-assembly of gold nanoparticles can be activated by enzymatic unveiling of surface-bound zwitterionic tetrapeptides through matrix metalloprotease-9 (MMP-9), which is overexpressed by cancer cells. This robust nanoparticle assembly is achieved by multivalent, self-complementary interactions of the zwitterionic tetrapeptides. In cancer cells that overexpress MMP-9, the nanoparticle assembly process occurs near the cell membrane and causes size-induced selection of cellular uptake mechanism, resulting in diminished cell growth. The enzyme responsiveness, and therefore, indirectly, the uptake route of the system can be programmed by customizing the peptide sequence: a simple inversion of the two amino acids at the cleavage site completely inactivates the enzyme responsiveness, self-assembly, and consequently changes the endocytic pathway. This robust self-complementary, zwitterionic peptide design demonstrates the use of enzyme-activated electrostatic side-chain patterns as powerful and customizable peptide modalities to program nanoparticle self-assembly and alter cellular response in biological context.},
  author       = {Huang, Richard H. and Nayeem, Nazia and He, Ye and Morales, Jorge and Graham, Duncan and Klajn, Rafal and Contel, Maria and O'Brien, Stephen and Ulijn, Rein V.},
  issn         = {1521-4095},
  journal      = {Advanced Materials},
  keywords     = {Mechanical Engineering, Mechanics of Materials, General Materials Science},
  number       = {1},
  publisher    = {Wiley},
  title        = {{Self‐complementary zwitterionic peptides direct nanoparticle assembly and enable enzymatic selection of endocytic pathways}},
  doi          = {10.1002/adma.202104962},
  volume       = {34},
  year         = {2022},
}

@article{13445,
  abstract     = {Rotation is typically assumed to induce strictly symmetric rotational splitting into the rotational multiplets of pure p- and g-modes. However, for evolved stars exhibiting mixed modes, avoided crossings between different multiplet components are known to yield asymmetric rotational splitting, in particular for near-degenerate mixed-mode pairs, where notional pure p-modes are fortuitously in resonance with pure g-modes. These near-degeneracy effects have been described in subgiants, but their consequences for the characterization of internal rotation in red giants have not previously been investigated in detail, in part owing to theoretical intractability. We employ new developments in the analytic theory of mixed-mode coupling to study these near-resonance phenomena. In the vicinity of the most p-dominated mixed modes, the near-degenerate intrinsic asymmetry from pure rotational splitting increases dramatically over the course of stellar evolution, and it depends strongly on the mode-mixing fraction ζ. We also find that a linear treatment of rotation remains viable for describing the underlying p- and g-modes, even when it does not for the resulting mixed modes undergoing these avoided crossings. We explore observational consequences for potential measurements of asymmetric mixed-mode splitting, which has been proposed as a magnetic-field diagnostic. Finally, we propose improved measurement techniques for rotational characterization, exploiting the linearity of rotational effects on the underlying p/g-modes, while still accounting for these mixed-mode coupling effects.},
  author       = {Ong, J. M. Joel and Bugnet, Lisa Annabelle and Basu, Sarbani},
  issn         = {1538-4357},
  journal      = {The Astrophysical Journal},
  keywords     = {Space and Planetary Science, Astronomy and Astrophysics},
  number       = {1},
  publisher    = {American Astronomical Society},
  title        = {{Mode mixing and rotational splittings. I. Near-degeneracy effects revisited}},
  doi          = {10.3847/1538-4357/ac97e7},
  volume       = {940},
  year         = {2022},
}

@article{13451,
  abstract     = {We characterize massive stars (M > 8 M⊙) in the nearby (D ∼ 0.8 Mpc) extremely metal-poor (Z ∼ 5% Z⊙) galaxy Leo A using Hubble Space Telescope ultraviolet (UV), optical, and near-infrared (NIR) imaging along with Keck/Low-Resolution Imaging Spectrograph and MMT/Binospec optical spectroscopy for 18 main-sequence OB stars. We find that: (a) 12 of our 18 stars show emission lines, despite not being associated with an H ii region, suggestive of stellar activity (e.g., mass loss, accretion, binary star interaction), which is consistent with previous predictions of enhanced activity at low metallicity; (b) six are Be stars, which are the first to be spectroscopically studied at such low metallicity—these Be stars have unusual panchromatic SEDs; (c) for stars well fit by the TLUSTY nonlocal thermodynamic equilibrium models, the photometric and spectroscopic values of $\mathrm{log}({T}_{\mathrm{eff}})$ and $\mathrm{log}(g)$ agree to within ∼0.01 dex and ∼0.18 dex, respectively, indicating that near-UV/optical/NIR imaging can be used to reliably characterize massive (M ∼ 8–30 M⊙) main-sequence star properties relative to optical spectroscopy; (d) the properties of the most-massive stars in H II regions are consistent with constraints from previous nebular emission line studies; and (e) 13 stars with M > 8M⊙ are >40 pc from a known star cluster or H II region. Our sample comprises ∼50% of all known massive stars at Z ≲ 10% Z⊙with derived stellar parameters, high-quality optical spectra, and panchromatic photometry.},
  author       = {Gull, Maude and Weisz, Daniel R. and Senchyna, Peter and Sandford, Nathan R. and Choi, Yumi and McLeod, Anna F. and El-Badry, Kareem and Götberg, Ylva Louise Linsdotter and Gilbert, Karoline M. and Boyer, Martha and Dalcanton, Julianne J. and GuhaThakurta, Puragra and Goldman, Steven and Marigo, Paola and McQuinn, Kristen B. W. and Pastorelli, Giada and Stark, Daniel P. and Skillman, Evan and Ting, Yuan-sen and Williams, Benjamin F.},
  issn         = {1538-4357},
  journal      = {The Astrophysical Journal},
  keywords     = {Space and Planetary Science, Astronomy and Astrophysics},
  number       = {2},
  publisher    = {American Astronomical Society},
  title        = {{A panchromatic study of massive stars in the extremely metal-poor local group dwarf galaxy Leo A}},
  doi          = {10.3847/1538-4357/aca295},
  volume       = {941},
  year         = {2022},
}

@article{13452,
  abstract     = {Magnetic fields can drastically change predictions of evolutionary models of massive stars via mass-loss quenching, magnetic braking, and efficient angular momentum transport, which we aim to quantify in this work. We use the MESA software instrument to compute an extensive main-sequence grid of stellar structure and evolution models, as well as isochrones, accounting for the effects attributed to a surface fossil magnetic field. The grid is densely populated in initial mass (3–60 M⊙), surface equatorial magnetic field strength (0–50 kG), and metallicity (representative of the Solar neighbourhood and the Magellanic Clouds). We use two magnetic braking and two chemical mixing schemes and compare the model predictions for slowly rotating, nitrogen-enriched (‘Group 2’) stars with observations in the Large Magellanic Cloud. We quantify a range of initial field strengths that allow for producing Group 2 stars and find that typical values (up to a few kG) lead to solutions. Between the subgrids, we find notable departures in surface abundances and evolutionary paths. In our magnetic models, chemical mixing is always less efficient compared to non-magnetic models due to the rapid spin-down. We identify that quasi-chemically homogeneous main sequence evolution by efficient mixing could be prevented by fossil magnetic fields. We recommend comparing this grid of evolutionary models with spectropolarimetric and spectroscopic observations with the goals of (i) revisiting the derived stellar parameters of known magnetic stars, and (ii) observationally constraining the uncertain magnetic braking and chemical mixing schemes.},
  author       = {Keszthelyi, Z and de Koter, A and Götberg, Ylva Louise Linsdotter and Meynet, G and Brands, S A and Petit, V and Carrington, M and David-Uraz, A and Geen, S T and Georgy, C and Hirschi, R and Puls, J and Ramalatswa, K J and Shultz, M E and ud-Doula, A},
  issn         = {1365-2966},
  journal      = {Monthly Notices of the Royal Astronomical Society},
  keywords     = {Space and Planetary Science, Astronomy and Astrophysics},
  number       = {2},
  pages        = {2028--2055},
  publisher    = {Oxford University Press},
  title        = {{The effects of surface fossil magnetic fields on massive star evolution: IV. Grids of models at Solar, LMC, and SMC metallicities}},
  doi          = {10.1093/mnras/stac2598},
  volume       = {517},
  year         = {2022},
}

@article{13992,
  abstract     = {Understanding the chirality of molecular reaction pathways is essential for a broad range of fundamental and applied sciences. However, the current ability to probe chirality on the time scale of primary processes underlying chemical reactions remains very limited. Here, we demonstrate time-resolved photoelectron circular dichroism (TRPECD) with ultrashort circularly polarized vacuum-ultraviolet (VUV) pulses from a tabletop source. We demonstrate the capabilities of VUV-TRPECD by resolving the chirality changes in time during the photodissociation of atomic iodine from two chiral molecules. We identify several general key features of TRPECD, which include the ability to probe dynamical chirality along the complete photochemical reaction path, the sensitivity to the local chirality of the evolving scattering potential, and the influence of electron scattering off dissociating photofragments. Our results are interpreted by comparison with high-level ab-initio calculations of transient PECDs from molecular photoionization calculations. Our experimental and theoretical techniques define a general approach to femtochirality.},
  author       = {Svoboda, Vít and Ram, Niraghatam Bhargava and Baykusheva, Denitsa Rangelova and Zindel, Daniel and Waters, Max D. J. and Spenger, Benjamin and Ochsner, Manuel and Herburger, Holger and Stohner, Jürgen and Wörner, Hans Jakob},
  issn         = {2375-2548},
  journal      = {Science Advances},
  keywords     = {Multidisciplinary},
  number       = {28},
  publisher    = {American Association for the Advancement of Science},
  title        = {{Femtosecond photoelectron circular dichroism of chemical reactions}},
  doi          = {10.1126/sciadv.abq2811},
  volume       = {8},
  year         = {2022},
}

@article{13993,
  abstract     = {Photoionization is a process taking place on attosecond time scales. How its properties evolve from isolated particles to the condensed phase is an open question of both fundamental and practical relevance. Here, we review recent work that has advanced the study of photoionization dynamics from atoms to molecules, clusters and the liquid phase. The first measurements of molecular photoionization delays have revealed the attosecond dynamics of electron emission from a molecular shape resonance and their sensitivity to the molecular potential. Using electron-ion coincidence spectroscopy these measurements have been extended from isolated molecules to clusters. A continuous increase of the delays with the water-cluster size has been observed up to a size of 4-5 molecules, followed by a saturation towards larger clusters. Comparison with calculations has revealed a correlation of the time delay with the spatial extension of the created electron hole. Using cylindrical liquid-microjet techniques, these measurements have also been extended to liquid water, revealing a delay relative to isolated water molecules that was very similar to the largest water clusters studied. Detailed modeling based on Monte-Carlo simulations confirmed that these delays are dominated by the contributions of the first two solvation shells, which agrees with the results of the cluster measurements. These combined results open the perspective of experimentally characterizing the delocalization of electronic wave functions in complex systems and studying their evolution on attosecond time scales.},
  author       = {Gong, Xiaochun and Jordan, Inga and Huppert, Martin and Heck, Saijoscha and Baykusheva, Denitsa Rangelova and Jelovina, Denis and Schild, Axel and Wörner, Hans Jakob},
  issn         = {2673-2424},
  journal      = {Chimia},
  keywords     = {General Medicine, General Chemistry},
  number       = {6},
  pages        = {520--528},
  publisher    = {Swiss Chemical Society},
  title        = {{Attosecond photoionization dynamics: from molecules over clusters to the liquid phase}},
  doi          = {10.2533/chimia.2022.520},
  volume       = {76},
  year         = {2022},
}

@article{13994,
  abstract     = {Ultrafast lasers are an increasingly important tool to control and stabilize emergent phases in quantum materials. Among a variety of possible excitation protocols, a particularly intriguing route is the direct light engineering of microscopic electronic parameters, such as the electron hopping and the local Coulomb repulsion (Hubbard 
U). In this work, we use time-resolved x-ray absorption spectroscopy to demonstrate the light-induced renormalization of the Hubbard U in a cuprate superconductor, La1.905Ba0.095CuO4. We show that intense femtosecond laser pulses induce a substantial redshift of the upper Hubbard band while leaving the Zhang-Rice singlet energy unaffected. By comparing the experimental data to time-dependent spectra of single- and three-band Hubbard models, we assign this effect to an approximately 140-meV reduction of the on-site Coulomb repulsion on the copper sites. Our demonstration of a dynamical Hubbard U renormalization in a copper oxide paves the way to a novel strategy for the manipulation of superconductivity and magnetism as well as to the realization of other long-range-ordered phases in light-driven quantum materials.},
  author       = {Baykusheva, Denitsa Rangelova and Jang, Hoyoung and Husain, Ali A. and Lee, Sangjun and TenHuisen, Sophia F. R. and Zhou, Preston and Park, Sunwook and Kim, Hoon and Kim, Jin-Kwang and Kim, Hyeong-Do and Kim, Minseok and Park, Sang-Youn and Abbamonte, Peter and Kim, B. J. and Gu, G. D. and Wang, Yao and Mitrano, Matteo},
  issn         = {2160-3308},
  journal      = {Physical Review X},
  keywords     = {General Physics and Astronomy},
  number       = {1},
  publisher    = {American Physical Society},
  title        = {{Ultrafast renormalization of the on-site Coulomb repulsion in a cuprate superconductor}},
  doi          = {10.1103/physrevx.12.011013},
  volume       = {12},
  year         = {2022},
}

@inproceedings{14093,
  abstract     = { We propose a stochastic conditional gradient method (CGM) for minimizing convex finite-sum objectives formed as a sum of smooth and non-smooth terms. Existing CGM variants for this template either suffer from slow convergence rates, or require carefully increasing the batch size over the course of the algorithm’s execution, which leads to computing full gradients. In contrast, the proposed method, equipped with a stochastic average gradient (SAG) estimator, requires only one sample per iteration. Nevertheless, it guarantees fast convergence rates on par with more sophisticated variance reduction techniques. In applications we put special emphasis on problems with a large number of separable constraints. Such problems are prevalent among semidefinite programming (SDP) formulations arising in machine learning and theoretical computer science. We provide numerical experiments on matrix completion, unsupervised clustering, and sparsest-cut SDPs. },
  author       = {Dresdner, Gideon and Vladarean, Maria-Luiza and Rätsch, Gunnar and Locatello, Francesco and Cevher, Volkan and Yurtsever, Alp},
  booktitle    = {Proceedings of the 25th International Conference on Artificial Intelligence and Statistics},
  issn         = {2640-3498},
  location     = {Virtual},
  pages        = {8439--8457},
  publisher    = {ML Research Press},
  title        = {{ Faster one-sample stochastic conditional gradient method for composite convex minimization}},
  volume       = {151},
  year         = {2022},
}

@article{14098,
  abstract     = {Magnetic fields can drastically change predictions of evolutionary models of massive stars via mass-loss quenching, magnetic braking, and efficient angular momentum transport, which we aim to quantify in this work. We use the MESA software instrument to compute an extensive main-sequence grid of stellar structure and evolution models, as well as isochrones, accounting for the effects attributed to a surface fossil magnetic field. The grid is densely populated in initial mass (3–60 M⊙), surface equatorial magnetic field strength (0–50 kG), and metallicity (representative of the Solar neighbourhood and the Magellanic Clouds). We use two magnetic braking and two chemical mixing schemes and compare the model predictions for slowly rotating, nitrogen-enriched (‘Group 2’) stars with observations in the Large Magellanic Cloud. We quantify a range of initial field strengths that allow for producing Group 2 stars and find that typical values (up to a few kG) lead to solutions. Between the subgrids, we find notable departures in surface abundances and evolutionary paths. In our magnetic models, chemical mixing is always less efficient compared to non-magnetic models due to the rapid spin-down. We identify that quasi-chemically homogeneous main sequence evolution by efficient mixing could be prevented by fossil magnetic fields. We recommend comparing this grid of evolutionary models with spectropolarimetric and spectroscopic observations with the goals of (i) revisiting the derived stellar parameters of known magnetic stars, and (ii) observationally constraining the uncertain magnetic braking and chemical mixing schemes.},
  author       = {Keszthelyi, Z. and Koter, A. de and Götberg, Ylva Louise Linsdotter and Meynet, G. and Brands, S. A. and Petit, V. and Carrington, M. and A. David-Uraz, A. David-Uraz and Geen, S. T. and Georgy, C. and Hirschi, R. and Puls, J. and Ramalatswa, K. J. and Shultz, M. E. and A. ud-Doula, A. ud-Doula},
  issn         = {1365-2966},
  journal      = {Monthly Notices of the Royal Astronomical Society},
  number       = {2},
  pages        = {2028--2055},
  publisher    = {Oxford Academic},
  title        = {{The effects of surface fossil magnetic fields on massive star evolution: IV. Grids of models at solar, LMC, and SMC metallicities}},
  doi          = {10.1093/mnras/stac2598},
  volume       = {517},
  year         = {2022},
}

@unpublished{14099,
  abstract     = {Magnetism can greatly impact the evolution of stars. In some stars with OBA spectral types there is direct evidence via the Zeeman effect for stable, large-scale magnetospheres, which lead to the spin-down of the stellar surface and reduced mass loss. So far, a comprehensive grid of stellar structure and evolution models accounting for these effects was lacking. For this reason, we computed and studied models with two magnetic braking and two chemical mixing schemes in three metallicity environments with the MESA software instrument. We find notable differences between the subgrids, which affects the model predictions and thus the detailed characterisation of stars. We are able to quantify the impact of magnetic fields in terms of preventing quasi-chemically homogeneous evolution and producing slowly-rotating, nitrogen-enriched ("Group 2") stars. Our model grid is fully open access and open source.},
  author       = {Keszthelyi, Z. and Koter, A. de and Götberg, Ylva Louise Linsdotter and Meynet, G. and Brands, S. A. and Petit, V. and Carrington, M. and A. David-Uraz, A. David-Uraz and Geen, S. T. and Georgy, C. and Hirschi, R. and Puls, J. and Ramalatswa, K. J. and Shultz, M. E. and A. ud-Doula, A. ud-Doula},
  booktitle    = {arXiv},
  title        = {{Spin-down and reduced mass loss in early-type stars with large-scale magnetic fields}},
  doi          = {10.48550/arXiv.2211.07060},
  year         = {2022},
}

@inproceedings{14106,
  abstract     = {We show that deep networks trained to satisfy demographic parity often do so
through a form of race or gender awareness, and that the more we force a network
to be fair, the more accurately we can recover race or gender from the internal state
of the network. Based on this observation, we investigate an alternative fairness
approach: we add a second classification head to the network to explicitly predict
the protected attribute (such as race or gender) alongside the original task. After
training the two-headed network, we enforce demographic parity by merging the
two heads, creating a network with the same architecture as the original network.
We establish a close relationship between existing approaches and our approach
by showing (1) that the decisions of a fair classifier are well-approximated by our
approach, and (2) that an unfair and optimally accurate classifier can be recovered
from a fair classifier and our second head predicting the protected attribute. We use
our explicit formulation to argue that the existing fairness approaches, just as ours,
demonstrate disparate treatment and that they are likely to be unlawful in a wide
range of scenarios under US law.},
  author       = {Lohaus, Michael and Kleindessner, Matthäus and Kenthapadi, Krishnaram and Locatello, Francesco and Russell, Chris},
  booktitle    = {36th Conference on Neural Information Processing Systems},
  isbn         = {9781713871088},
  location     = {New Orleans, LA, United States},
  pages        = {16548--16562},
  publisher    = {Neural Information Processing Systems Foundation},
  title        = {{Are two heads the same as one? Identifying disparate treatment in fair neural networks}},
  volume       = {35},
  year         = {2022},
}

@inproceedings{14107,
  abstract     = {Amodal perception requires inferring the full shape of an object that is partially occluded. This task is particularly challenging on two levels: (1) it requires more information than what is contained in the instant retina or imaging sensor, (2) it is difficult to obtain enough well-annotated amodal labels for supervision. To this end, this paper develops a new framework of
Self-supervised amodal Video object segmentation (SaVos). Our method efficiently leverages the visual information of video temporal sequences to infer the amodal mask of objects. The key intuition is that the occluded part of an object can be explained away if that part is visible in other frames, possibly deformed as long as the deformation can be reasonably learned.
Accordingly, we derive a novel self-supervised learning paradigm that efficiently utilizes the visible object parts as the supervision to guide the training on videos. In addition to learning type prior to complete masks for known types, SaVos also learns the spatiotemporal prior, which is also useful for the amodal task and could generalize to unseen types. The proposed
framework achieves the state-of-the-art performance on the synthetic amodal segmentation benchmark FISHBOWL and the real world benchmark KINS-Video-Car. Further, it lends itself well to being transferred to novel distributions using test-time adaptation, outperforming existing models even after the transfer to a new distribution.},
  author       = {Yao, Jian and Hong, Yuxin and Wang, Chiyu and Xiao, Tianjun and He, Tong and Locatello, Francesco and Wipf, David and Fu, Yanwei and Zhang, Zheng},
  booktitle    = {36th Conference on Neural Information Processing Systems},
  location     = {New Orleans, LA, United States},
  title        = {{Self-supervised amodal video object segmentation}},
  doi          = {10.48550/arXiv.2210.12733},
  year         = {2022},
}

@inproceedings{14114,
  abstract     = {Algorithmic fairness is frequently motivated in terms of a trade-off in which overall performance is decreased so as to improve performance on disadvantaged groups where the algorithm would otherwise be less accurate. Contrary to this, we find that applying existing fairness approaches to computer vision improve fairness by degrading the performance of classifiers across all groups (with increased degradation on the best performing groups). Extending the bias-variance decomposition for classification to fairness, we theoretically explain why the majority of fairness methods designed for low capacity models should not be used in settings involving high-capacity models, a scenario common to computer vision. We corroborate this analysis with extensive experimental support that shows that many of the fairness heuristics used in computer vision also degrade performance on the most disadvantaged groups. Building on these insights, we propose an adaptive augmentation strategy that, uniquely, of all methods tested, improves performance for the disadvantaged groups.},
  author       = {Zietlow, Dominik and Lohaus, Michael and Balakrishnan, Guha and Kleindessner, Matthaus and Locatello, Francesco and Scholkopf, Bernhard and Russell, Chris},
  booktitle    = {2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  isbn         = {9781665469470},
  issn         = {2575-7075},
  location     = {New Orleans, LA, United States},
  pages        = {10400--10411},
  publisher    = {Institute of Electrical and Electronics Engineers},
  title        = {{Leveling down in computer vision: Pareto inefficiencies in fair deep classifiers}},
  doi          = {10.1109/cvpr52688.2022.01016},
  year         = {2022},
}

@inproceedings{14168,
  abstract     = {Recent work has seen the development of general purpose neural architectures
that can be trained to perform tasks across diverse data modalities. General
purpose models typically make few assumptions about the underlying
data-structure and are known to perform well in the large-data regime. At the
same time, there has been growing interest in modular neural architectures that
represent the data using sparsely interacting modules. These models can be more
robust out-of-distribution, computationally efficient, and capable of
sample-efficient adaptation to new data. However, they tend to make
domain-specific assumptions about the data, and present challenges in how
module behavior (i.e., parameterization) and connectivity (i.e., their layout)
can be jointly learned. In this work, we introduce a general purpose, yet
modular neural architecture called Neural Attentive Circuits (NACs) that
jointly learns the parameterization and a sparse connectivity of neural modules
without using domain knowledge. NACs are best understood as the combination of
two systems that are jointly trained end-to-end: one that determines the module
configuration and the other that executes it on an input. We demonstrate
qualitatively that NACs learn diverse and meaningful module configurations on
the NLVR2 dataset without additional supervision. Quantitatively, we show that
by incorporating modularity in this way, NACs improve upon a strong non-modular
baseline in terms of low-shot adaptation on CIFAR and CUBs dataset by about
10%, and OOD robustness on Tiny ImageNet-R by about 2.5%. Further, we find that
NACs can achieve an 8x speedup at inference time while losing less than 3%
performance. Finally, we find NACs to yield competitive results on diverse data
modalities spanning point-cloud classification, symbolic processing and
text-classification from ASCII bytes, thereby confirming its general purpose
nature.},
  author       = {Rahaman, Nasim and Weiss, Martin and Locatello, Francesco and Pal, Chris and Bengio, Yoshua and Schölkopf, Bernhard and Li, Li Erran and Ballas, Nicolas},
  booktitle    = {36th Conference on Neural Information Processing Systems},
  location     = {New Orleans, United States},
  title        = {{Neural attentive circuits}},
  volume       = {35},
  year         = {2022},
}

@inproceedings{14170,
  abstract     = {The idea behind object-centric representation learning is that natural scenes can better be modeled as compositions of objects and their relations as opposed to distributed representations. This inductive bias can be injected into neural networks to potentially improve systematic generalization and performance of downstream tasks in scenes with multiple objects. In this paper, we train state-of-the-art unsupervised models on five common multi-object datasets and evaluate segmentation metrics and downstream object property prediction. In addition, we study generalization and robustness by investigating the settings where either a single object is out of distribution -- e.g., having an unseen color, texture, or shape -- or global properties of the scene are altered -- e.g., by occlusions, cropping, or increasing the number of objects. From our experimental study, we find object-centric representations to be useful for
downstream tasks and generally robust to most distribution shifts affecting objects. However, when the distribution shift affects the input in a less structured manner, robustness in terms of segmentation and downstream task performance may vary significantly across models and distribution shifts. },
  author       = {Dittadi, Andrea and Papa, Samuele and Vita, Michele De and Schölkopf, Bernhard and Winther, Ole and Locatello, Francesco},
  booktitle    = {Proceedings of the 39th International Conference on Machine Learning},
  location     = {Baltimore, MD, United States},
  pages        = {5221--5285},
  publisher    = {ML Research Press},
  title        = {{Generalization and robustness implications in object-centric learning}},
  volume       = {2022},
  year         = {2022},
}

@inproceedings{14171,
  abstract     = {This paper demonstrates how to recover causal graphs from the score of the
data distribution in non-linear additive (Gaussian) noise models. Using score
matching algorithms as a building block, we show how to design a new generation
of scalable causal discovery methods. To showcase our approach, we also propose
a new efficient method for approximating the score's Jacobian, enabling to
recover the causal graph. Empirically, we find that the new algorithm, called
SCORE, is competitive with state-of-the-art causal discovery methods while
being significantly faster.},
  author       = {Rolland, Paul and Cevher, Volkan and Kleindessner, Matthäus and Russel, Chris and Schölkopf, Bernhard and Janzing, Dominik and Locatello, Francesco},
  booktitle    = {Proceedings of the 39th International Conference on Machine Learning},
  location     = {Baltimore, MD, United States},
  pages        = {18741--18753},
  publisher    = {ML Research Press},
  title        = {{Score matching enables causal discovery of nonlinear additive noise  models}},
  volume       = {162},
  year         = {2022},
}

@inproceedings{14172,
  abstract     = {An important component for generalization in machine learning is to uncover underlying latent factors of variation as well as the mechanism through which each factor acts in the world. In this paper, we test whether 17 unsupervised, weakly supervised, and fully supervised representation learning approaches correctly infer the generative factors of variation in simple datasets (dSprites, Shapes3D, MPI3D) from controlled environments, and on our contributed CelebGlow dataset. In contrast to prior robustness work that introduces novel factors of variation during test time, such as blur or other (un)structured noise, we here recompose, interpolate, or extrapolate only existing factors of variation from the training data set (e.g., small and medium-sized objects during training and large objects during testing). Models
that learn the correct mechanism should be able to generalize to this benchmark. In total, we train and test 2000+ models and observe that all of them struggle to learn the underlying mechanism regardless of supervision signal and architectural bias. Moreover, the generalization capabilities of all tested models drop significantly as we move from artificial datasets towards
more realistic real-world datasets. Despite their inability to identify the correct mechanism, the models are quite modular as their ability to infer other in-distribution factors remains fairly stable, providing only a single factoris out-of-distribution. These results point to an important yet understudied problem of learning mechanistic models of observations that can facilitate
generalization.},
  author       = {Schott, Lukas and Kügelgen, Julius von and Träuble, Frederik and Gehler, Peter and Russell, Chris and Bethge, Matthias and Schölkopf, Bernhard and Locatello, Francesco and Brendel, Wieland},
  booktitle    = {10th International Conference on Learning Representations},
  location     = {Virtual},
  title        = {{Visual representation learning does not generalize strongly within the  same domain}},
  year         = {2022},
}

@inproceedings{14173,
  abstract     = {Since out-of-distribution generalization is a generally ill-posed problem, various proxy targets (e.g., calibration, adversarial robustness, algorithmic corruptions, invariance across shifts) were studied across different research programs resulting in different recommendations. While sharing the same aspirational goal, these approaches have never been tested under the same
experimental conditions on real data. In this paper, we take a unified view of previous work, highlighting message discrepancies that we address empirically, and providing recommendations on how to measure the robustness of a model and how to improve it. To this end, we collect 172 publicly available dataset pairs for training and out-of-distribution evaluation of accuracy, calibration error, adversarial attacks, environment invariance, and synthetic corruptions. We fine-tune over 31k networks, from nine different architectures in the many- and
few-shot setting. Our findings confirm that in- and out-of-distribution accuracies tend to increase jointly, but show that their relation is largely dataset-dependent, and in general more nuanced and more complex than posited by previous, smaller scale studies.},
  author       = {Wenzel, Florian and Dittadi, Andrea and Gehler, Peter Vincent and Carl-Johann Simon-Gabriel, Carl-Johann Simon-Gabriel and Horn, Max and Zietlow, Dominik and Kernert, David and Russell, Chris and Brox, Thomas and Schiele, Bernt and Schölkopf, Bernhard and Locatello, Francesco},
  booktitle    = {36th Conference on Neural Information Processing Systems},
  isbn         = {9781713871088},
  location     = {New Orleans, LA, United States},
  pages        = {7181--7198},
  publisher    = {Neural Information Processing Systems Foundation},
  title        = {{Assaying out-of-distribution generalization in transfer learning}},
  volume       = {35},
  year         = {2022},
}