@article{17860, abstract = {Radicals are unique molecular systems for applications in electronic devices due to their open-shell electronic structures. Radicals can function as good electrical conductors and switches in molecular circuits while also holding great promise in the field of molecular spintronics. However, it is both challenging to create stable, persistent radicals and to understand their properties in molecular junctions. The goal of this Perspective is to address this dual challenge by providing design principles for the synthesis of stable radicals relevant to molecular junctions, as well as offering current insight into the electronic properties of radicals in single-molecule devices. By exploring both the chemical and physical properties of established radical systems, we will facilitate increased exploration and development of radical-based molecular systems.}, author = {Li, Liang and Prindle, Claudia R. and Shi, Wanzhuo and Nuckolls, Colin and Venkataraman, Latha}, issn = {1520-5126}, journal = {Journal of the American Chemical Society}, number = {33}, pages = {18182--18204}, publisher = {American Chemical Society}, title = {{Radical single-molecule junctions}}, doi = {10.1021/jacs.3c04487}, volume = {145}, year = {2023}, } @article{17861, abstract = {Molecular one-dimensional topological insulators (1D TIs), described by the Su-Schrieffer-Heeger (SSH) model, are a new class of molecular electronic wires whose low-energy topological edge states endow them with high electrical conductivity. However, when these 1D TIs become long, the high conductance is not sustained because the coupling between the edge states decreases with increasing length. Here, we present a new design where we connect multiple short 1D SSH TI units linearly or in a cycle to create molecular wires with a continuous topological state density. Using a tight-binding method, we show that the linear system gives a length-independent conductance. The cyclic systems show an interesting odd-even effect, with unit transmission in the topological limit, but zero transmission in the trivial limit. Furthermore, based on our calculations, we predict that these systems can support resonant transmission with a quantum of conductance. We can further expand these results to phenylene-based linear and cyclic 1D TI systems and confirm the length-dependent conductance in such systems. }, author = {Li, Liang and Nuckolls, Colin and Venkataraman, Latha}, issn = {1948-7185}, journal = {The Journal of Physical Chemistry Letters}, number = {22}, pages = {5141--5147}, publisher = {American Chemical Society}, title = {{Designing long and highly conducting molecular wires with multiple nontrivial topological states}}, doi = {10.1021/acs.jpclett.3c01081}, volume = {14}, year = {2023}, } @article{17862, abstract = {Electric field acceleration of alkyl hydroperoxide activation to acylate amines in the scanning tunneling microscope-based break-junction is reported. Alkyl hydroperoxide mixtures, generated from hydrocarbon autoxidation in air, were found to be competent reagents for the functionalization of gold surfaces. Intermolecular coupling on the surface in the presence of amines was observed, yielding normal alkylamides. This novel mode of alkyl hydroperoxide activation to generate acylium equivalents was found to be responsive to the magnitude of the bias in the break junction, indicating an electric field influence on this novel reactivity.}, author = {Wang, Xiye and Zhang, Boyuan and Fowler, Brandon and Venkataraman, Latha and Rovis, Tomislav}, issn = {1520-5126}, journal = {Journal of the American Chemical Society}, number = {22}, pages = {11903--11906}, publisher = {American Chemical Society}, title = {{Alkane solvent-derived acylation reaction driven by electric fields}}, doi = {10.1021/jacs.3c02064}, volume = {145}, year = {2023}, } @article{17863, abstract = {Understanding and tuning charge transport over a single molecule is a fundamental topic in molecular electronics. Single-molecule junctions composed of individual molecules attached to two electrodes are the most common components built for single-molecule charge transport studies. During the past two decades, rapid technical and theoretical advances in single-molecule junctions have increased our understanding of the conductance properties and functions of molecular devices. In this perspective article, we introduce the basic principles of charge transport in single-molecule junctions, then give an overview of recent progress in modulating single-molecule transport through external stimuli such as electric field and potential, light, mechanical force, heat, and chemical environment. Lastly, we discuss challenges and offer views on future developments in molecular electronics.}, author = {Zou, Qi and Qiu, Jin and Zang, Yaping and Tian, He and Venkataraman, Latha}, issn = {2667-1417}, journal = {eScience}, number = {3}, publisher = {Elsevier BV}, title = {{Modulating single-molecule charge transport through external stimulus}}, doi = {10.1016/j.esci.2023.100115}, volume = {3}, year = {2023}, } @article{17864, abstract = {Molecular one-dimensional topological insulators (1D TIs), which conduct through energetically low-lying topological edge states, can be extremely highly conducting and exhibit a reversed conductance decay, affording them great potential as building blocks for nanoelectronic devices. However, these properties can only be observed at the short length limit. To extend the length at which these anomalous effects can be observed, we design topological oligo[n]emeraldine wires using short 1D TIs as building blocks. As the wire length increases, the number of topological states increases, enabling an increased electronic transmission along the wire; specifically, we show that we can drive over a microampere current through a single ∼5 nm molecular wire, appreciably more than what has been observed in other long wires reported to date. Calculations and experiments show that the longest oligo[7]emeraldine with doped topological states has over 106 enhancements in the transmission compared to its pristine form. The discovery of these highly conductive, long organic wires helps overcome a fundamental hurdle to implementing molecules in complex, nanoscale circuitry: their structures become too insulating at lengths that are useful in designing nanoscale circuits.}, author = {Li, Liang and Louie, Shayan and Evans, Austin M. and Meirzadeh, Elena and Nuckolls, Colin and Venkataraman, Latha}, issn = {1520-5126}, journal = {Journal of the American Chemical Society}, number = {4}, pages = {2492--2498}, publisher = {American Chemical Society}, title = {{Topological radical pairs produce ultrahigh conductance in long molecular wires}}, doi = {10.1021/jacs.2c12059}, volume = {145}, year = {2023}, } @article{17865, abstract = {Understanding how molecular geometry affects the electronic properties of single-molecule junctions experimentally has been challenging. Typically, metal–molecule–metal junctions are measured using a break-junction method where electrode separation is mechanically evolving during measurement. Here, to probe the impact of the junction geometry on conductance, we apply a sinusoidal modulation to the molecular junction electrode position. Simultaneously, we probe the nonlinearity of the current–voltage characteristics of each junction through a modulation in the applied bias at a different frequency. In turn, we show that junctions formed with molecules that have different molecule–electrode interfaces exhibit statistically distinguishable Fourier-transformed conductances. In particular, we find a marked bias dependence for the modulation of junctions where transmission is mediated thorough the van der Waals (vdW) interaction. We attribute our findings to voltage-modulated vdW interactions at the single-molecule level.}, author = {Wei, Yujing and Li, Liang and Greenwald, Julia E. and Venkataraman, Latha}, issn = {1530-6992}, journal = {Nano Letters}, number = {2}, pages = {567--572}, publisher = {American Chemical Society}, title = {{Voltage-modulated van der waals interaction in single-molecule junctions}}, doi = {10.1021/acs.nanolett.2c04098}, volume = {23}, year = {2023}, } @article{17866, abstract = {Electric fields have been used to control and direct chemical reactions in biochemistry and enzymatic catalysis, yet directly applying external electric fields to activate reactions in bulk solution and to characterize them ex situ remains a challenge. Here we utilize the scanning tunneling microscope-based break-junction technique to investigate the electric field driven homolytic cleavage of the radical initiator 4-(methylthio)benzoic peroxyanhydride at ambient temperatures in bulk solution, without the use of co-initiators or photochemical activators. Through time-dependent ex situ quantification by high performance liquid chromatography using a UV-vis detector, we find that the electric field catalyzes the reaction. Importantly, we demonstrate that the reaction rate in a field increases linearly with the solvent dielectric constant. Using density functional theory calculations, we show that the applied electric field decreases the dissociation energy of the O–O bond and stabilizes the product relative to the reactant due to their different dipole moments.}, author = {Zhang, Boyuan and Schaack, Cedric and Prindle, Claudia R. and Vo, Ethan A. and Aziz, Miriam and Steigerwald, Michael L. and Berkelbach, Timothy C. and Nuckolls, Colin and Venkataraman, Latha}, issn = {2041-6539}, journal = {Chemical Science}, number = {7}, pages = {1769--1774}, publisher = {Royal Society of Chemistry}, title = {{Electric fields drive bond homolysis}}, doi = {10.1039/d2sc06411a}, volume = {14}, year = {2023}, } @article{18179, abstract = {Linnik type problems concern the distribution of projections of integral points on the unit sphere as their norm increases, and different generalizations of this phenomenon. Our work addresses a question of this type: we prove the uniform distribution of the projections of primitive Z2 points in the p-adic unit sphere, as their (real) norm tends to infinity. The proof is via counting lattice points in semi-simple S-arithmetic groups.}, author = {Guilloux, Antonin and Horesh, Tal}, issn = {2592-6616}, journal = {Publications mathématiques de Besançon - Algèbre et Théorie des nombres}, pages = {85--107}, publisher = {Presses Universitaires de Franche-Comté}, title = {{p-adic directions of primitive vectors}}, doi = {10.5802/pmb.50}, volume = {2023}, year = {2023}, } @article{18189, abstract = {Strongly interacting topological matter1 exhibits fundamentally new phenomena with potential applications in quantum information technology2,3. Emblematic instances are fractional quantum Hall (FQH) states4, in which the interplay of a magnetic field and strong interactions gives rise to fractionally charged quasi-particles, long-ranged entanglement and anyonic exchange statistics. Progress in engineering synthetic magnetic fields5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21 has raised the hope to create these exotic states in controlled quantum systems. However, except for a recent Laughlin state of light22, preparing FQH states in engineered systems remains elusive. Here we realize a FQH state with ultracold atoms in an optical lattice. The state is a lattice version of a bosonic ν = 1/2 Laughlin state4,23 with two particles on 16 sites. This minimal system already captures many hallmark features of Laughlin-type FQH states24,25,26,27,28: we observe a suppression of two-body interactions, we find a distinctive vortex structure in the density correlations and we measure a fractional Hall conductivity of σH/σ0 = 0.6(2) by means of the bulk response to a magnetic perturbation. Furthermore, by tuning the magnetic field, we map out the transition point between the normal and the FQH regime through a spectroscopic investigation of the many-body gap. Our work provides a starting point for exploring highly entangled topological matter with ultracold atoms29,30,31,32,33.}, author = {Leonard, Julian and Kim, Sooshin and Kwan, Joyce and Segura, Perrin and Grusdt, Fabian and Repellin, Cécile and Goldman, Nathan and Greiner, Markus}, issn = {1476-4687}, journal = {Nature}, number = {7970}, pages = {495--499}, publisher = {Springer Nature}, title = {{Realization of a fractional quantum Hall state with ultracold atoms}}, doi = {10.1038/s41586-023-06122-4}, volume = {619}, year = {2023}, } @article{18190, abstract = {Strongly correlated systems can exhibit unexpected phenomena when brought in a state far from equilibrium. An example is many-body localization, which prevents generic interacting systems from reaching thermal equilibrium even at long times1,2. The stability of the many-body localized phase has been predicted to be hindered by the presence of small thermal inclusions that act as a bath, leading to the delocalization of the entire system through an avalanche propagation mechanism3,4,5,6,7,8. Here we study the dynamics of a thermal inclusion of variable size when it is coupled to a many-body localized system. We find evidence for accelerated transport of thermal inclusion into the localized region. We monitor how the avalanche spreads through the localized system and thermalizes it site by site by measuring the site-resolved entropy over time. Furthermore, we isolate the strongly correlated bath-induced dynamics with multipoint correlations between the bath and the system. Our results have implications on the robustness of many-body localized systems and their critical behaviour.}, author = {Leonard, Julian and Kim, Sooshin and Rispoli, Matthew and Lukin, Alexander and Schittko, Robert and Kwan, Joyce and Demler, Eugene and Sels, Dries and Greiner, Markus}, issn = {1745-2481}, journal = {Nature Physics}, number = {4}, pages = {481--485}, publisher = {Springer Nature}, title = {{Probing the onset of quantum avalanches in a many-body localized system}}, doi = {10.1038/s41567-022-01887-3}, volume = {19}, year = {2023}, } @article{18207, abstract = {Comparison of myoglobin structures reveals that protein isolated from horse heart consistently adopts an alternate turn conformation in comparison to its homologues. Analysis of hundreds of high-resolution structures discounts crystallization conditions or the surrounding amino acid protein environment as explaining this difference, that is also not captured by the AlphaFold prediction. Rather, a water molecule is identified as stabilizing the conformation in the horse heart structure, which immediately reverts to the whale conformation in molecular dynamics simulations excluding that structural water.}, author = {Bronstein, Alexander and Marx, Ailie}, issn = {2045-2322}, journal = {Scientific Reports}, publisher = {Springer Nature}, title = {{Water stabilizes an alternate turn conformation in horse heart myoglobin}}, doi = {10.1038/s41598-023-32821-z}, volume = {13}, year = {2023}, } @article{18208, abstract = {The holy grail of materials science is de novo molecular design, meaning engineering molecules with desired characteristics. The introduction of generative deep learning has greatly advanced efforts in this direction, yet molecular discovery remains challenging and often inefficient. Herein we introduce GaUDI, a guided diffusion model for inverse molecular design that combines an equivariant graph neural net for property prediction and a generative diffusion model. We demonstrate GaUDI’s effectiveness in designing molecules for organic electronic applications by using single- and multiple-objective tasks applied to a generated dataset of 475,000 polycyclic aromatic systems. GaUDI shows improved conditional design, generating molecules with optimal properties and even going beyond the original distribution to suggest better molecules than those in the dataset. In addition to point-wise targets, GaUDI can also be guided toward open-ended targets (for example, a minimum or maximum) and in all cases achieves close to 100% validity of generated molecules.}, author = {Weiss, Tomer and Mayo Yanes, Eduardo and Chakraborty, Sabyasachi and Cosmo, Luca and Bronstein, Alexander and Gershoni-Poranne, Renana}, issn = {2662-8457}, journal = {Nature Computational Science}, number = {10}, pages = {873--882}, publisher = {Springer Nature}, title = {{Guided diffusion for inverse molecular design}}, doi = {10.1038/s43588-023-00532-0}, volume = {3}, year = {2023}, } @article{18209, abstract = {In this work, interpretable deep learning was used to identify structure–property relationships governing the HOMO–LUMO gap and the relative stability of polybenzenoid hydrocarbons (PBHs) using a ring-based graph representation. This representation was combined with a subunit-based perception of PBHs, allowing chemical insights to be presented in terms of intuitive and simple structural motifs. The resulting insights agree with conventional organic chemistry knowledge and electronic structure-based analyses and also reveal new behaviors and identify influential structural motifs. In particular, we evaluated and compared the effects of linear, angular, and branching motifs on these two molecular properties and explored the role of dispersion in mitigating the torsional strain inherent in nonplanar PBHs. Hence, the observed regularities and the proposed analysis contribute to a deeper understanding of the behavior of PBHs and form the foundation for design strategies for new functional PBHs.}, author = {Weiss, Tomer and Wahab, Alexandra and Bronstein, Alexander and Gershoni-Poranne, Renana}, issn = {1520-6904}, journal = {The Journal of Organic Chemistry}, number = {14}, pages = {9645--9656}, publisher = {American Chemical Society}, title = {{Interpretable deep-learning unveils structure–property relationships in polybenzenoid hydrocarbons}}, doi = {10.1021/acs.joc.2c02381}, volume = {88}, year = {2023}, } @inproceedings{18212, abstract = {The high memory bandwidth demand of sparse embedding layers continues to be a critical challenge in scaling the performance of recommendation models. While prior works have exploited heterogeneous memory system designs and partial embedding sum memoization techniques, they offer limited benefits. This is because prior designs either target a very small subset of embeddings to simplify their analysis or incur a high processing cost to account for all embeddings, which does not scale with the large sizes of modern embedding tables. This paper proposes GRACE-a lightweight and scalable graph-based algorithm-system co-design framework to significantly improve the embedding layer performance of recommendation models. GRACE proposes a novel Item Co-occurrence Graph (ICG) that scalably records item co-occurrences. GRACE then presents a new system-aware ICG clustering algorithm to find frequently accessed item combinations of arbitrary lengths to compute and memoize their partial sums. High-frequency partial sums are stored in a software-managed cache space to reduce memory traffic and improve the throughput of computing sparse features. We further present a cache data layout and low-cost address computation logic to efficiently lookup item embeddings and their partial sums. Our evaluation shows that GRACE significantly outperforms the state-of-the-art techniques SPACE and MERCI by 1.5x and 1.4x, respectively.}, author = {Ye, Haojie and Vedula, Sanketh and Chen, Yuhan and Yang, Yichen and Bronstein, Alexander and Dreslinski, Ronald and Mudge, Trevor and Talati, Nishil}, booktitle = {Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems}, isbn = {9781450399180}, number = {3}, pages = {282--301}, publisher = {Association for Computing Machinery}, title = {{GRACE: A scalable graph-based approach to accelerating recommendation model inference}}, doi = {10.1145/3582016.3582029}, volume = {11}, year = {2023}, } @article{18213, abstract = {What is the best way to match the nodes of two graphs? This graph alignment problem generalizes graph isomorphism and arises in applications from social network analysis to bioinformatics. Some solutions assume that auxiliary information on known matches or node or edge attributes is available, or utilize arbitrary graph features. Such methods fare poorly in the pure form of the problem, in which only graph structures are given. Other proposals translate the problem to one of aligning node embeddings, yet, by doing so, provide only a single-scale view of the graph. In this article, we transfer the shape-analysis concept of functional maps from the continuous to the discrete case, and treat the graph alignment problem as a special case of the problem of finding a mapping between functions on graphs. We present GRASP, a method that first establishes a correspondence between functions derived from Laplacian matrix eigenvectors, which capture multiscale structural characteristics, and then exploits this correspondence to align nodes. We enhance the basic form of GRASP by altering two of its components, namely the embedding method and the assignment procedure it employs, leveraging its modular, hence adaptable design. Our experimental study, featuring noise levels higher than anything used in previous studies, shows that the enhanced form of GRASP outperforms scalable state-of-the-art methods for graph alignment across noise levels and graph types, and performs competitively with respect to the best non-scalable ones. We include in our study another modular graph alignment algorithm, CONE, which is also adaptable thanks to its modular nature, and show it can manage graphs with skewed power-law degree distributions.}, author = {Hermanns, Judith and Skitsas, Konstantinos and Tsitsulin, Anton and Munkhoeva, Marina and Kyster, Alexander and Nielsen, Simon and Bronstein, Alexander and Mottin, Davide and Karras, Panagiotis}, issn = {1556-472X}, journal = {ACM Transactions on Knowledge Discovery from Data}, number = {4}, publisher = {Association for Computing Machinery}, title = {{GRASP: Scalable graph alignment by spectral corresponding functions}}, doi = {10.1145/3561058}, volume = {17}, year = {2023}, } @article{18214, abstract = {Graph sparsification is a technique that approximates a given graph by a sparse graph with a subset of vertices and/or edges. The goal of an effective sparsification algorithm is to maintain specific graph properties relevant to the downstream task while minimizing the graph's size. Graph algorithms often suffer from long execution time due to the irregularity and the large real-world graph size. Graph sparsification can be applied to greatly reduce the run time of graph algorithms by substituting the full graph with a much smaller sparsified graph, without significantly degrading the output quality. However, the interaction between numerous sparsifiers and graph properties is not widely explored, and the potential of graph sparsification is not fully understood. In this work, we cover 16 widely-used graph metrics, 12 representative graph sparsification algorithms, and 14 real-world input graphs spanning various categories, exhibiting diverse characteristics, sizes, and densities. We developed a framework to extensively assess the performance of these sparsification algorithms against graph metrics, and provide insights to the results. Our study shows that there is no one sparsifier that performs the best in preserving all graph properties, e.g. sparsifiers that preserve distance-related graph properties (eccentricity) struggle to perform well on Graph Neural Networks (GNN). This paper presents a comprehensive experimental study evaluating the performance of sparsification algorithms in preserving essential graph metrics. The insights inform future research in incorporating matching graph sparsification to graph algorithms to maximize benefits while minimizing quality degradation. Furthermore, we provide a framework to facilitate the future evaluation of evolving sparsification algorithms, graph metrics, and ever-growing graph data.}, author = {Chen, Yuhan and Ye, Haojie and Vedula, Sanketh and Bronstein, Alexander and Dreslinski, Ronald and Mudge, Trevor and Talati, Nishil}, issn = {2150-8097}, journal = {Proceedings of the VLDB Endowment}, number = {3}, pages = {427--440}, publisher = {Association for Computing Machinery}, title = {{Demystifying graph sparsification algorithms in graph properties preservation}}, doi = {10.14778/3632093.3632106}, volume = {17}, year = {2023}, } @inproceedings{18215, abstract = {We study the problem of real-time scheduling in a multi-hop millimeter-wave (mmWave) mesh. We develop a model-free deep reinforcement learning algorithm called Adaptive Activator RL (AARL), which determines the subset of mmWave links that should be activated during each time slot and the power level for each link. The most important property of AARL is its ability to make scheduling decisions within the strict time frame constraints of typical 5G mmWave networks. AARL can handle a variety of network topologies, network loads, and interference models, it can also adapt to different workloads. We demonstrate the operation of AARL on several topologies: a small topology with 10 links, a moderately-sized mesh with 48 links, and a large topology with 96 links. We show that for each topology, we compare the throughput obtained by AARL to that of a benchmark algorithm called RPMA (Residual Profit Maximizer Algorithm). The most important advantage of AARL compared to RPMA is that it is much faster and can make the necessary scheduling decisions very rapidly during every time slot, while RPMA cannot. In addition, the quality of the scheduling decisions made by AARL outperforms those made by RPMA.}, author = {Gahtan, Barak and Cohen, Reuven and Bronstein, Alexander and Kedar, Gil}, booktitle = {14th International Conference on Network of the Future}, isbn = {9798350338089}, issn = {2833-0072}, location = {Izmir, Turkiye}, pages = {71--79}, publisher = {IEEE}, title = {{Using deep reinforcement learning for mmWave real-time scheduling}}, doi = {10.1109/nof58724.2023.10302794}, year = {2023}, } @article{18216, abstract = {Protein structure, both at the global and local level, dictates function. Proteins fold from chains of amino acids, forming secondary structures, α-helices and β-strands, that, at least for globular proteins, subsequently fold into a three-dimensional structure. Here, we show that a Ramachandran-type plot focusing on the two dihedral angles separated by the peptide bond, and entirely contained within an amino acid pair, defines a local structural unit. We further demonstrate the usefulness of this cross-peptide-bond Ramachandran plot by showing that it captures β-turn conformations in coil regions, that traditional Ramachandran plot outliers fall into occupied regions of our plot, and that thermophilic proteins prefer specific amino acid pair conformations. Further, we demonstrate experimentally that the effect of a point mutation on backbone conformation and protein stability depends on the amino acid pair context, i.e., the identity of the adjacent amino acid, in a manner predictable by our method.}, author = {Rosenberg, Aviv A. and Yehishalom, Nitsan and Marx, Ailie and Bronstein, Alexander}, issn = {1091-6490}, journal = {Proceedings of the National Academy of Sciences}, number = {44}, publisher = {National Academy of Sciences}, title = {{An amino-domino model described by a cross-peptide-bond Ramachandran plot defines amino acid pairs as local structural units}}, doi = {10.1073/pnas.2301064120}, volume = {120}, year = {2023}, } @inproceedings{18217, abstract = {A central challenge in building robotic prostheses is the creation of a sensor-based system able to read physiological signals from the lower limb and instruct a robotic hand to perform various tasks. Existing systems typically perform discrete gestures such as pointing or grasping, by employing electromyography (EMG) or ultrasound (US) technologies to analyze muscle states. While estimating finger gestures has been done in the past by detecting prominent gestures, we are interested in detection, or inference, done in the context of fine motions that evolve over time. Examples include motions occurring when performing fine and dexterous tasks such as keyboard typing or piano playing. We consider this task as an important step towards higher adoption rates of robotic prostheses among arm amputees, as it has the potential to dramatically increase functionality in performing daily tasks. To this end, we present an end-to-end robotic system, which can successfully infer fine finger motions. This is achieved by modeling the hand as a robotic manipulator and using it as an intermediate representation to encode muscles' dynamics from a sequence of US images. We evaluated our method by collecting data from a group of subjects and demonstrating how it can be used to replay music played or text typed. To the best of our knowledge, this is the first study demonstrating these downstream tasks within an end-to-end system.}, author = {Zadok, Dean and Salzman, Oren and Wolf, Alon and Bronstein, Alexander}, booktitle = {2023 IEEE International Conference on Robotics and Automation}, location = {London, United Kingdom}, publisher = {IEEE}, title = {{Towards predicting fine finger motions from ultrasound images via kinematic representation}}, doi = {10.1109/icra48891.2023.10160601}, volume = {27}, year = {2023}, } @inproceedings{18218, abstract = {Deep neural networks are known to be susceptible to adversarial perturbations – small perturbations that alter the output of the network and exist under strict norm limitations. While such perturbations are usually discussed as tailored to a specific input, a universal perturbation can be constructed to alter the model’s output on a set of inputs. Universal perturbations present a more realistic case of adversarial attacks, as awareness of the model’s exact input is not required. In addition, the universal attack setting raises the subject of generalization to unseen data, where given a set of inputs, the universal perturbations aim to alter the model’s output on out-of-sample data. In this work, we study physical passive patch adversarial attacks on visual odometry-based autonomous navigation systems. A visual odometry system aims to infer the relative camera motion between two corresponding viewpoints, and is frequently used by vision-based autonomous navigation systems to estimate their state. For such navigation systems, a patch adversarial perturbation poses a severe security issue, as it can be used to mislead a system onto some collision course. To the best of our knowledge, we show for the first time that the error margin of a visual odometry model can be significantly increased by deploying patch adversarial attacks in the scene. We provide evaluation on synthetic closed-loop drone navigation data and demonstrate that a comparable vulnerability exists in real data. A reference implementation of the proposed method and the reported experiments is provided at https://github.com/patchadversarialattacks/patchadversarialattacks.}, author = {Nemcovsky, Yaniv and Jacoby, Matan and Bronstein, Alexander and Baskin, Chaim}, booktitle = {16th Asian Conference on Computer Vision}, isbn = {9783031262920}, issn = {1611-3349}, location = {Macao, China}, pages = {518--534}, publisher = {Springer Nature}, title = {{Physical passive patch adversarial attacks on visual odometry systems}}, doi = {10.1007/978-3-031-26293-7_31}, volume = {13847}, year = {2023}, }