@article{17608,
abstract = {We study the long-term evolution of the global structure of axisymmetric accretion flows onto a black hole (BH) at rates substantially higher than the Eddington value (M˙Edd), performing two-dimensional hydrodynamical simulations with and without radiative diffusion. In the high-accretion optically-thick limit, where the radiation energy is efficiently trapped within the inflow, the accretion flow becomes adiabatic and comprises of turbulent gas in the equatorial region and strong bipolar outflows. As a result, the mass inflow rate decreases toward the center as M˙in∝rp with p∼0.5−0.7 and a small fraction of the inflowing gas feeds the nuclear BH. Thus, super-Eddington accretion is sustained only when a larger amount of gas is supplied from larger radii at >100−1000 M˙Edd. The global structure of the flow settles down to a quasi-steady state in millions of the orbital timescale at the BH event horizon, which is >10−100 times longer than that addressed in previous (magneto-)RHD simulation studies. Energy transport via radiative diffusion accelerates the outflow near the poles in the inner region but does not change the overall properties of the accretion flow compared to the cases without diffusion. Based on our simulation results, we provide a mechanical feedback model for super-Eddington accreting BHs. This can be applied as a sub-grid model in large-scale cosmological simulations that do not sufficiently resolve galactic nuclei, and to the formation of the heaviest gravitational-wave sources via accretion in dense environments.},
author = {Hu, Haojie and Inayoshi, Kohei and Haiman, Zoltán and Quataert, Eliot and Kuiper, Rolf},
issn = {0004-637X},
journal = {The Astrophysical Journal},
number = {2},
publisher = {American Astronomical Society},
title = {{Long-term evolution of supercritical black hole accretion with outflows: A subgrid feedback model for cosmological simulations}},
doi = {10.3847/1538-4357/ac75d8},
volume = {934},
year = {2022},
}
@article{17868,
abstract = {Reversed conductance decay describes increasing conductance of a molecular chain series with increasing chain length. Realizing reversed conductance decay is an important step toward making long and highly conducting molecular wires. Recent work has shown that one-dimensional topological insulators (1D TIs) can exhibit reversed conductance decay due to their nontrivial edge states. The Su–Schrieffer–Heeger (SSH) model for 1D TIs relates to the electronic structure of these isolated molecules but not their electron transport properties as single-molecule junctions. Herein, we use a tight-binding approach to demonstrate that polyacetylene and other diradicaloid 1D TIs show a reversed conductance decay at the short chain limit. We explain these conductance trends by analyzing the impact of the edge states in these 1D systems on the single-molecule junction transmission. Additionally, we discuss how the self-energy from the electrode-molecule coupling and the on-site energy of the edge sites can be tuned to create longer wires with reversed conductance decays.},
author = {Li, Liang and Gunasekaran, Suman and Wei, Yujing and Nuckolls, Colin and Venkataraman, Latha},
issn = {1948-7185},
journal = {The Journal of Physical Chemistry Letters},
number = {41},
pages = {9703--9710},
publisher = {American Chemical Society},
title = {{Reversed conductance decay of 1D topological insulators by tight-binding analysis}},
doi = {10.1021/acs.jpclett.2c02812},
volume = {13},
year = {2022},
}
@article{17869,
abstract = {The formation of carbon–carbon bonds with transition metal reagents serves as a cornerstone of organic synthesis. Here, we show that the reactivity of an otherwise kinetically inert transition metal complex can be induced by an external electric field to affect a coupling reaction. These results highlight the importance of electric field effects in reaction chemistry and offers a new strategy to modulate organometallic reactivity.},
author = {Orchanian, Nicholas M. and Guizzo, Sophia and Steigerwald, Michael L. and Nuckolls, Colin and Venkataraman, Latha},
issn = {1364-548X},
journal = {Chemical Communications},
number = {90},
pages = {12556--12559},
publisher = {Royal Society of Chemistry},
title = {{Electric-field-induced coupling of aryl iodides with a nickel(0) complex}},
doi = {10.1039/d2cc03671a},
volume = {58},
year = {2022},
}
@article{17870,
abstract = {The electric fields created at solid–liquid interfaces are important in heterogeneous catalysis. Here we describe the Ullmann coupling of aryl iodides on rough gold surfaces, which we monitor in situ using the scanning tunneling microscope-based break junction (STM-BJ) and ex situ using mass spectrometry and fluorescence spectroscopy. We find that this Ullmann coupling reaction occurs only on rough gold surfaces in polar solvents, the latter of which implicates interfacial electric fields. These experimental observations are supported by density functional theory calculations that elucidate the roles of surface roughness and local electric fields on the reaction. More broadly, this touchstone study offers a facile method to access and probe in real time an increasingly prominent yet incompletely understood mode of catalysis.},
author = {Stone, Ilana B. and Starr, Rachel L. and Hoffmann, Norah and Wang, Xiao and Evans, Austin M. and Nuckolls, Colin and Lambert, Tristan H. and Steigerwald, Michael L. and Berkelbach, Timothy C. and Roy, Xavier and Venkataraman, Latha},
issn = {2041-6539},
journal = {Chemical Science},
number = {36},
pages = {10798--10805},
publisher = {Royal Society of Chemistry},
title = {{Interfacial electric fields catalyze Ullmann coupling reactions on gold surfaces}},
doi = {10.1039/d2sc03780g},
volume = {13},
year = {2022},
}
@article{17871,
abstract = {Single-molecule topological insulators are promising candidates as conducting wires over nanometre length scales. A key advantage is their ability to exhibit quasi-metallic transport, in contrast to conjugated molecular wires which typically exhibit a low conductance that decays as the wire length increases. Here, we study a family of oligophenylene-bridged bis(triarylamines) with tunable and stable mono- or di-radicaloid character. These wires can undergo one- and two-electron chemical oxidations to the corresponding mono-cation and di-cation, respectively. We show that the oxidized wires exhibit reversed conductance decay with increasing length, consistent with the expectation for Su–Schrieffer–Heeger-type one-dimensional topological insulators. The 2.6-nm-long di-cation reported here displays a conductance greater than 0.1G0, where G0 is the conductance quantum, a factor of 5,400 greater than the neutral form. The observed conductance–length relationship is similar between the mono-cation and di-cation series. Density functional theory calculations elucidate how the frontier orbitals and delocalization of radicals facilitate the observed non-classical quasi-metallic behaviour.},
author = {Li, Liang and Low, Jonathan Z. and Wilhelm, Jan and Liao, Guanming and Gunasekaran, Suman and Prindle, Claudia R. and Starr, Rachel L. and Golze, Dorothea and Nuckolls, Colin and Steigerwald, Michael L. and Evers, Ferdinand and Campos, Luis M. and Yin, Xiaodong and Venkataraman, Latha},
issn = {1755-4349},
journal = {Nature Chemistry},
number = {9},
pages = {1061--1067},
publisher = {Springer Nature},
title = {{Highly conducting single-molecule topological insulators based on mono- and di-radical cations}},
doi = {10.1038/s41557-022-00978-1},
volume = {14},
year = {2022},
}
@article{17872,
abstract = {Coherent tunneling electron transport through molecular wires has been theoretically established as a temperature-independent process. Although several experimental studies have shown counter examples, robust models to describe this temperature dependence have not been thoroughly developed. Here, we demonstrate that dynamic molecular structures lead to temperature-dependent conductance within coherent tunneling regime. Using a custom-built variable-temperature scanning tunneling microscopy break-junction instrument, we find that oligo[n]phenylenes exhibit clear temperature-dependent conductance. Our calculations reveal that thermally activated dihedral rotations allow these molecular wires to have a higher probability of being in a planar conformation. As the tunneling occurs primarily through π-orbitals, enhanced coplanarization substantially increases the time-averaged tunneling probability. These calculations are consistent with the observation that more rotational pivot points in longer molecular wires leads to larger temperature-dependence on conductance. These findings reveal that molecular conductance within coherent and off-resonant electron transport regimes can be controlled by manipulating dynamic molecular structure.},
author = {Lee, Woojung and Louie, Shayan and Evans, Austin M. and Orchanian, Nicholas M. and Stone, Ilana B. and Zhang, Boyuan and Wei, Yujing and Roy, Xavier and Nuckolls, Colin and Venkataraman, Latha},
issn = {1530-6992},
journal = {Nano Letters},
number = {12},
pages = {4919--4924},
publisher = {American Chemical Society},
title = {{Increased molecular conductance in Oligo[n]phenylene wires by thermally enhanced dihedral planarization}},
doi = {10.1021/acs.nanolett.2c01549},
volume = {22},
year = {2022},
}
@article{17873,
abstract = {AbstractA critical overview of the theory of the chirality‐induced spin selectivity (CISS) effect, that is, phenomena in which the chirality of molecular species imparts significant spin selectivity to various electron processes, is provided. Based on discussions in a recently held workshop, and further work published since, the status of CISS effects—in electron transmission, electron transport, and chemical reactions—is reviewed. For each, a detailed discussion of the state‐of‐the‐art in theoretical understanding is provided and remaining challenges and research opportunities are identified.},
author = {Evers, Ferdinand and Aharony, Amnon and Bar‐Gill, Nir and Entin‐Wohlman, Ora and Hedegård, Per and Hod, Oded and Jelinek, Pavel and Kamieniarz, Grzegorz and Lemeshko, Mikhail and Michaeli, Karen and Mujica, Vladimiro and Naaman, Ron and Paltiel, Yossi and Refaely‐Abramson, Sivan and Tal, Oren and Thijssen, Jos and Thoss, Michael and van Ruitenbeek, Jan M. and Venkataraman, Latha and Waldeck, David H. and Yan, Binghai and Kronik, Leeor},
issn = {0935-9648},
journal = {Advanced Materials},
number = {13},
publisher = {Wiley},
title = {{Theory of chirality induced spin selectivity: Progress and challenges}},
doi = {10.1002/adma.202106629},
volume = {34},
year = {2022},
}
@article{17874,
abstract = {Redox-active two-dimensional polymers (RA-2DPs) are promising lithium battery organic cathode materials due to their regular porosities and high chemical stabilities. However, weak electrical conductivities inherent to the non-conjugated molecular motifs used thus far limit device performance and the practical relevance of these materials. We herein address this problem by developing a modular approach to construct π-conjugated RA-2DPs with a new polycyclic aromatic redox-active building block PDI-DA. Efficient imine-condensation between PDI-DA and two polyfunctional amine nodes followed by quantitative alkyl chain removal produced RA-2DPs TAPPy-PDI and TAPB-PDI as conjugated, porous, polycrystalline networks. In-plane conjugation and permanent porosity endow these materials with high electrical conductivity and high ion diffusion rates. As such, both RA-2DPs function as organic cathode materials with good rate performance and excellent cycling stability. Importantly, the improved design enables higher areal mass-loadings than were previously available, which drives a practical demonstration of TAPPy-PDI as the power source for a series of LED lights. Collectively, this investigation discloses viable synthetic methodologies and design principles for the realization of high-performance organic cathode materials.},
author = {Jin, Zexin and Cheng, Qian and Evans, Austin M. and Gray, Jesse and Zhang, Ruiwen and Bao, Si Tong and Wei, Fengkai and Venkataraman, Latha and Yang, Yuan and Nuckolls, Colin},
issn = {2041-6539},
journal = {Chemical Science},
number = {12},
pages = {3533--3538},
publisher = {Royal Society of Chemistry},
title = {{π-Conjugated redox-active two-dimensional polymers as organic cathode materials}},
doi = {10.1039/d1sc07157b},
volume = {13},
year = {2022},
}
@article{17875,
abstract = {Nanoscale plasmonic structures have been primarily characterized through scattering studies, but electroluminescence offers an exciting alternative from a technological standpoint by removing the need for optical excitation. In sub-nanometer biased junctions, electronic tunneling can serve as the excitation source for plasmon-coupled electroluminescence, but the gap size dependence to this plasmonic enhancement has not been characterized. Here, we simultaneously probe the electroluminescence and conductance of Au tunnel junctions. We find that plasmonic enhancement increases as the gap size is reduced for junctions biased between 1.4 and 1.8 V, consistent with the behavior of charge transfer plasmons. At biases above 1.9 V, we see decreasing plasmonic enhancement with the decreasing gap, showing quenching due to tunneling in remarkable agreement with the trends observed for high energy plasmons in scattering experiments. Critically, we find that plasmonic enhancement of electroluminescence is gap size-dependent and, furthermore, is in agreement with the nature of modes excited by scattering.},
author = {Paoletta, Angela L. and Fung, E-Dean and Venkataraman, Latha},
issn = {2330-4022},
journal = {ACS Photonics},
number = {2},
pages = {688--693},
publisher = {American Chemical Society},
title = {{Gap size-dependent plasmonic enhancement in electroluminescent tunnel junctions}},
doi = {10.1021/acsphotonics.1c01757},
volume = {9},
year = {2022},
}
@article{18191,
abstract = {Large-scale quantum devices provide insights beyond the reach of classical simulations. However, for a reliable and verifiable quantum simulation, the building blocks of the quantum device require exquisite benchmarking. This benchmarking of large-scale dynamical quantum systems represents a major challenge due to lack of efficient tools for their simulation. Here, we present a scalable algorithm based on neural networks for Hamiltonian tomography in out-of-equilibrium quantum systems. We illustrate our approach using a model for a forefront quantum simulation platform: ultracold atoms in optical lattices. Specifically, we show that our algorithm is able to reconstruct the Hamiltonian of an arbitrary sized bosonic ladder system using an accessible amount of experimental measurements. We are able to significantly increase the previously known parameter precision.},
author = {Valenti, Agnes and Jin, Guliuxin and Leonard, Julian and Huber, Sebastian D. and Greplova, Eliska},
issn = {2469-9934},
journal = {Physical Review A},
number = {2},
publisher = {American Physical Society},
title = {{Scalable Hamiltonian learning for large-scale out-of-equilibrium quantum dynamics}},
doi = {10.1103/physreva.105.023302},
volume = {105},
year = {2022},
}
@article{18211,
abstract = {Deep neural networks are known to be vulnerable to malicious perturbations. Current methods for improving adversarial robustness make use of either implicit or explicit regularization, with the latter is usually based on adversarial training. Randomized smoothing, the averaging of the classifier outputs over a random distribution centered in the sample, has been shown to guarantee a classifier’s performance subject to bounded perturbations of the input. In this work, we study the application of randomized smoothing to improve performance on unperturbed data and increase robustness to adversarial attacks. We propose to combine smoothing along with adversarial training and randomization approaches, and find that doing so significantly improves the resilience compared to the baseline. We examine our method’s performance on common whitebox (FGSM, PGD) and black-box (transferable attack and NAttack) attacks on CIFAR-10 and CIFAR-100, and determine that for a low number of iterations, smoothing provides a significant performance boost that persists even for perturbations with a high attack norm, . For example, under a PGD-10 attack on CIFAR-10 using Wide-ResNet28-4, we achieve 60.3% accuracy for infinity norm ∞ = 8/255 and 13.1% accuracy for ∞ = 35/255 – outperforming previous art by 3% and 6%, respectively. We achieve nearly twice the accuracy on ∞ = 35/255 and even more so for perturbations with higher infinity norm. A reference implementation of the proposed method is provided. },
author = {Nemcovsky, Yaniv and Zheltonozhskii, Evgenii and Baskin, Chaim and Chmiel, Brian and Bronstein, Alexander and Mendelson, Avi},
issn = {1573-7497},
journal = {Applied Intelligence},
number = {8},
pages = {9483--9498},
publisher = {Springer Nature},
title = {{Adversarial robustness via noise injection in smoothed models}},
doi = {10.1007/s10489-022-03423-5},
volume = {53},
year = {2022},
}
@article{18220,
abstract = {Synonymous codons translate into the same amino acid. Although the identity of synonymous codons is often considered inconsequential to the final protein structure, there is mounting evidence for an association between the two. Our study examined this association using regression and classification models, finding that codon sequences predict protein backbone dihedral angles with a lower error than amino acid sequences, and that models trained with true dihedral angles have better classification of synonymous codons given structural information than models trained with random dihedral angles. Using this classification approach, we investigated local codon–codon dependencies and tested whether synonymous codon identity can be predicted more accurately from codon context than amino acid context alone, and most specifically which codon context position carries the most predictive power.},
author = {Ackerman-Schraier, Linor and Rosenberg, Aviv A. and Marx, Ailie and Bronstein, Alexander},
issn = {2045-2322},
journal = {Scientific Reports},
publisher = {Springer Nature},
title = {{Machine learning approaches demonstrate that protein structures carry information about their genetic coding}},
doi = {10.1038/s41598-022-25874-z},
volume = {12},
year = {2022},
}
@article{18221,
abstract = {Synonymous codons translate into chemically identical amino acids. Once considered inconsequential to the formation of the protein product, there is evidence to suggest that codon usage affects co-translational protein folding and the final structure of the expressed protein. Here we develop a method for computing and comparing codon-specific Ramachandran plots and demonstrate that the backbone dihedral angle distributions of some synonymous codons are distinguishable with statistical significance for some secondary structures. This shows that there exists a dependence between codon identity and backbone torsion of the translated amino acid. Although these findings cannot pinpoint the causal direction of this dependence, we discuss the vast biological implications should coding be shown to directly shape protein conformation and demonstrate the usefulness of this method as a tool for probing associations between codon usage and protein structure. Finally, we urge for the inclusion of exact genetic information into structural databases.},
author = {Rosenberg, Aviv A. and Marx, Ailie and Bronstein, Alexander},
issn = {2041-1723},
journal = {Nature Communications},
publisher = {Springer Nature},
title = {{Codon-specific Ramachandran plots show amino acid backbone conformation depends on identity of the translated codon}},
doi = {10.1038/s41467-022-30390-9},
volume = {13},
year = {2022},
}
@article{18222,
abstract = {STUDY QUESTION: What is the accuracy and agreement of embryologists when assessing the implantation probability of blastocysts using time-lapse imaging (TLI), and can it be improved with a data-driven algorithm?
SUMMARY ANSWER: The overall interobserver agreement of a large panel of embryologists was moderate and prediction accuracy was modest, while the purpose-built artificial intelligence model generally resulted in higher performance metrics.
WHAT IS KNOWN ALREADY: Previous studies have demonstrated significant interobserver variability amongst embryologists when assessing embryo quality. However, data concerning embryologists’ ability to predict implantation probability using TLI is still lacking. Emerging technologies based on data-driven tools have shown great promise for improving embryo selection and predicting clinical outcomes.
STUDY DESIGN, SIZE, DURATION: TLI video files of 136 embryos with known implantation data were retrospectively collected from two clinical sites between 2018 and 2019 for the performance assessment of 36 embryologists and comparison with a deep neural network (DNN).
PARTICIPANTS/MATERIALS, SETTING, METHODS: We recruited 39 embryologists from 13 different countries. All participants were blinded to clinical outcomes. A total of 136 TLI videos of embryos that reached the blastocyst stage were used for this experiment. Each embryo’s likelihood of successfully implanting was assessed by 36 embryologists, providing implantation probability grades (IPGs) from 1 to 5, where 1 indicates a very low likelihood of implantation and 5 indicates a very high likelihood. Subsequently, three embryologists with over 5 years of experience provided Gardner scores. All 136 blastocysts were categorized into three quality groups based on their Gardner scores. Embryologist predictions were then converted into predictions of implantation (IPG ≥ 3) and no implantation (IPG ≤ 2). Embryologists’ performance and agreement were assessed using Fleiss kappa coefficient. A 10-fold cross-validation DNN was developed to provide IPGs for TLI video files. The model’s performance was compared to that of the embryologists.
MAIN RESULTS AND THE ROLE OF CHANCE: Logistic regression was employed for the following confounding variables: country of residence, academic level, embryo scoring system, log years of experience and experience using TLI. None were found to have a statistically significant impact on embryologist performance at α = 0.05. The average implantation prediction accuracy for the embryologists was 51.9% for all embryos (N = 136). The average accuracy of the embryologists when assessing top quality and poor quality embryos (according to the Gardner score categorizations) was 57.5% and 57.4%, respectively, and 44.6% for fair quality embryos. Overall interobserver agreement was moderate (κ = 0.56, N = 136). The best agreement was achieved in the poor + top quality group (κ = 0.65, N = 77), while the agreement in the fair quality group was lower (κ = 0.25, N = 59). The DNN showed an overall accuracy rate of 62.5%, with accuracies of 62.2%, 61% and 65.6% for the poor, fair and top quality groups, respectively. The AUC for the DNN was higher than that of the embryologists overall (0.70 DNN vs 0.61 embryologists) as well as in all of the Gardner groups (DNN vs embryologists—Poor: 0.69 vs 0.62; Fair: 0.67 vs 0.53; Top: 0.77 vs 0.54).
LIMITATIONS, REASONS FOR CAUTION: Blastocyst assessment was performed using video files acquired from time-lapse incubators, where each video contained data from a single focal plane. Clinical data regarding the underlying cause of infertility and endometrial thickness before the transfer was not available, yet may explain implantation failure and lower accuracy of IPGs. Implantation was defined as the presence of a gestational sac, whereas the detection of fetal heartbeat is a more robust marker of embryo viability. The raw data were anonymized to the extent that it was not possible to quantify the number of unique patients and cycles included in the study, potentially masking the effect of bias from a limited patient pool. Furthermore, the lack of demographic data makes it difficult to draw conclusions on how representative the dataset was of the wider population. Finally, embryologists were required to assess the implantation potential, not embryo quality. Although this is not the traditional approach to embryo evaluation, morphology/morphokinetics as a means of assessing embryo quality is believed to be strongly correlated with viability and, for some methods, implantation potential.
WIDER IMPLICATIONS OF THE FINDINGS: Embryo selection is a key element in IVF success and continues to be a challenge. Improving the predictive ability could assist in optimizing implantation success rates and other clinical outcomes and could minimize the financial and emotional burden on the patient. This study demonstrates moderate agreement rates between embryologists, likely due to the subjective nature of embryo assessment. In particular, we found that average embryologist accuracy and agreement were significantly lower for fair quality embryos when compared with that for top and poor quality embryos. Using data-driven algorithms as an assistive tool may help IVF professionals increase success rates and promote much needed standardization in the IVF clinic. Our results indicate a need for further research regarding technological advancement in this field.},
author = {Fordham, Daniel E and Rosentraub, Dror and Polsky, Avital L and Aviram, Talia and Wolf, Yotam and Perl, Oriel and Devir, Asnat and Rosentraub, Shahar and Silver, David H and Gold Zamir, Yael and Bronstein, Alexander and Lara Lara, Miguel and Ben Nagi, Jara and Alvarez, Adrian and Munné, Santiago},
issn = {1460-2350},
journal = {Human Reproduction},
number = {10},
pages = {2275--2290},
publisher = {Oxford University Press},
title = {{Embryologist agreement when assessing blastocyst implantation probability: Is data-driven prediction the solution to embryo assessment subjectivity?}},
doi = {10.1093/humrep/deac171},
volume = {37},
year = {2022},
}
@inbook{18223,
abstract = {The term silent mutation is commonly used to describe (1) a change in the DNA sequence that does not result in an observable effect on the organism’s phenotype; and (2) a synonymous mutation where the nucleotide change leaves the translated amino acid sequence unchanged. When Christian Anfinsen showed that a folded and active protein could be denatured to lose structure and activity and then subsequently renatured to regain the same structure and activity it appeared that the native, thermodynamically stable, structure of a protein depends only on the amino acid sequence and solution conditions (Anfinsen and Haber 1961). This experiment suggested that, once translated, proteins carry no memory of the genetic sequence and led to one of the most erroneous assumptions in modern science; synonymous codons were long considered silent, a mutation of the type that has no effect on an organism’s phenotype.},
author = {Rosenberg, Aviv A. and Bronstein, Alexander and Marx, Ailie},
booktitle = {Single Nucleotide Polymorphisms},
editor = {Sauna, Zuben E. and Kimchi-Sarfaty, Chava},
isbn = {9783031056147},
pages = {37--47},
publisher = {Springer Nature},
title = {{Recording Silence – Accurate Annotation of the Genetic Sequence Is Required to Better Understand How Synonymous Coding Affects Protein Structure and Disease}},
doi = {10.1007/978-3-031-05616-1_3},
year = {2022},
}
@article{18224,
abstract = {Learning from one or few visual examples is one of the key capabilities of humans since early infancy, but is still a significant challenge for modern AI systems. While considerable progress has been achieved in few-shot learning from a few image examples, much less attention has been given to the verbal descriptions that are usually provided to infants when they are presented with a new object. In this paper, we focus on the role of additional semantics that can significantly facilitate few-shot visual learning. Building upon recent advances in few-shot learning with additional semantic information, we demonstrate that further improvements are possible by combining multiple and richer semantics (category labels, attributes, and natural language descriptions). Using these ideas, we offer the community new results on the popular miniImageNet and CUB few-shot benchmarks, comparing favorably to the previous state-of-the-art results for both visual only and visual plus semantics-based approaches. We also performed an ablation study investigating the components and design choices of our approach. Code available on github.com/EliSchwartz/mutiple-semantics.},
author = {Schwartz, Eli and Karlinsky, Leonid and Feris, Rogerio and Giryes, Raja and Bronstein, Alexander},
issn = {0167-8655},
journal = {Pattern Recognition Letters},
pages = {142--147},
publisher = {Elsevier},
title = {{Baby steps towards few-shot learning with multiple semantics}},
doi = {10.1016/j.patrec.2022.06.012},
volume = {160},
year = {2022},
}
@article{18225,
abstract = {Isometric feature mapping is an established time-honored algorithm in manifold learning and non-linear dimensionality reduction. Its prominence can be attributed to the output of a coherent global low-dimensional representation of data by preserving intrinsic distances. In order to enable an efficient and more applicable isometric feature mapping, a diverse set of sophisticated advancements have been proposed to the original algorithm to incorporate important factors like sparsity of computation, conformality, topological constraints and spectral geometry. However, a significant shortcoming of most approaches is the dependence on large-scale dense-spectral decompositions and the inability to generalize to points far away from the sampling of the manifold.
In this paper, we explore an unsupervised deep learning approach for computing distance-preserving maps for non-linear dimensionality reduction. We demonstrate that our framework is general enough to incorporate all previous advancements and show a significantly improved local and non-local generalization of the isometric mapping. Our approach involves training with only a few landmark points and avoids the need for population of dense matrices as well as computing their spectral decomposition.},
author = {Pai, Gautam and Bronstein, Alexander and Talmon, Ronen and Kimmel, Ron},
issn = {0262-8856},
journal = {Image and Vision Computing},
publisher = {Elsevier},
title = {{Deep isometric maps}},
doi = {10.1016/j.imavis.2022.104461},
volume = {123},
year = {2022},
}
@article{18226,
abstract = {Spontaneous parametric downconversion (SPDC) in quantum optics is an invaluable resource for the realization of high-dimensional qudits with spatial modes of light. One of the main open challenges is how to directly generate a desirable qudit state in the SPDC process. This problem can be addressed through advanced computational learning methods; however, due to difficulties in modeling the SPDC process by a fully differentiable algorithm, progress has been limited. Here, we overcome these limitations and introduce a physically constrained and differentiable model, validated against experimental results for shaped pump beams and structured crystals, capable of learning the relevant interaction parameters in the process. We avoid any restrictions induced by the stochastic nature of our physical model and integrate the dynamic equations governing the evolution under the SPDC Hamiltonian. We solve the inverse problem of designing a nonlinear quantum optical system that achieves the desired quantum state of downconverted photon pairs. The desired states are defined using either the second-order correlations between different spatial modes or by specifying the required density matrix. By learning nonlinear photonic crystal structures as well as different pump shapes, we successfully show how to generate maximally entangled states. Furthermore, we simulate all-optical coherent control over the generated quantum state by actively changing the profile of the pump beam. Our work can be useful for applications such as novel designs of high-dimensional quantum key distribution and quantum information processing protocols. In addition, our method can be readily applied for controlling other degrees of freedom of light in the SPDC process, such as spectral and temporal properties, and may even be used in condensed-matter systems having a similar interaction Hamiltonian.},
author = {Rozenberg, Eyal and Karnieli, Aviv and Yesharim, Ofir and Foley-Comer, Joshua and Trajtenberg-Mills, Sivan and Freedman, Daniel and Bronstein, Alexander and Arie, Ady},
issn = {2334-2536},
journal = {Optica},
number = {6},
pages = {602--615},
publisher = {Optica Publishing Group},
title = {{Inverse design of spontaneous parametric downconversion for generation of high-dimensional qudits}},
doi = {10.1364/optica.451115},
volume = {9},
year = {2022},
}
@article{18227,
abstract = {Existing cross-modal hashing methods ignore the informative multimodal joint information and cannot fully exploit the semantic labels. In this paper, we propose a deep fused two-step cross-modal hashing (DFTH) framework with multiple semantic supervision. In the first step, DFTH learns unified hash codes for instances by a fusion network. Semantic label and similarity reconstruction have been introduced to acquire binary codes that are informative, discriminative and semantic similarity preserving. In the second step, two modality-specific hash networks are learned under the supervision of common hash codes reconstruction, label reconstruction, and intra-modal and inter-modal semantic similarity reconstruction. The modality-specific hash networks can generate semantic preserving binary codes for out-of-sample queries. To deal with the vanishing gradients of binarization, continuous differentiable tanh is introduced to approximate the discrete sign function, making the networks able to back-propagate by automatic gradient computation. Extensive experiments on MIRFlickr25K and NUS-WIDE show the superiority of DFTH over state-of-the-art methods.},
author = {Kang, Peipei and Lin, Zehang and Yang, Zhenguo and Bronstein, Alexander and Li, Qing and Liu, Wenyin},
issn = {1573-7721},
journal = {Multimedia Tools and Applications},
number = {11},
pages = {15653--15670},
publisher = {Springer Nature},
title = {{Deep fused two-step cross-modal hashing with multiple semantic supervision}},
doi = {10.1007/s11042-022-12187-6},
volume = {81},
year = {2022},
}
@inproceedings{18229,
abstract = {We present Self-Classifier – a novel self-supervised end-to-end classification learning approach. Self-Classifier learns labels and representations simultaneously in a single-stage end-to-end manner by optimizing for same-class prediction of two augmented views of the same sample. To guarantee non-degenerate solutions (i.e., solutions where all labels are assigned to the same class) we propose a mathematically motivated variant of the cross-entropy loss that has a uniform prior asserted on the predicted labels. In our theoretical analysis, we prove that degenerate solutions are not in the set of optimal solutions of our approach. Self-Classifier is simple to implement and scalable. Unlike other popular unsupervised classification and contrastive representation learning approaches, it does not require any form of pre-training, expectation-maximization, pseudo-labeling, external clustering, a second network, stop-gradient operation, or negative pairs. Despite its simplicity, our approach sets a new state of the art for unsupervised classification of ImageNet; and even achieves comparable to state-of-the-art results for unsupervised representation learning. Code is available at https://github.com/elad-amrani/self-classifier.},
author = {Amrani, Elad and Karlinsky, Leonid and Bronstein, Alexander},
booktitle = {17th European Conference on Computer Vision},
isbn = {9783031198205},
issn = {1611-3349},
location = {Tel Aviv, Israel},
pages = {116--132},
publisher = {Springer Nature},
title = {{Self-supervised classification network}},
doi = {10.1007/978-3-031-19821-2_7},
volume = {13691},
year = {2022},
}