Search

We present Facade (Fast and Accurate Contextual Anomaly DEtection): a high-precision deep-learning-based anomaly detection system deployed at Google (a large technology company) as the last line of defense against insider threats since 2018. Facade is an innovative unsupervised action-context system that detects suspicious actions by considering the context surrounding each action, including relevant facts about the user and other entities involved. It is built around a new multi-modal model that is trained on corporate document access, SQL query, and HTTP/RPC request logs. To overcome the scarcity of incident data, Facade harnesses a novel contrastive learning strategy that relies solely on benign data. Its use of history and implicit social network featurization efficiently handles the frequent out-of-distribution events that occur in a rapidly changing corporate environment, and sustains Facade's high precision performance for a full year after training. Beyond the core model, Facade contributes an innovative clustering approach based on user and action embeddings to improve detection robustness and achieve high precision, multi-scale detection. Functionally what sets Facade apart from existing anomaly detection systems is its high precision. It detects insider attackers with an extremely low false positive rate, lower than 0.01%. For single rogue actions, such as the illegitimate access to a sensitive document, the false positive rate is as low as 0.0003%. To the best of our knowledge, Facade is the only published insider risk anomaly detection system that helps secure such a large corporate environment., Comment: Under review

Non-alcoholic fatty liver disease (NAFLD) is one of the most widespread liver disorders on a global scale, posing a significant threat of progressing to more severe conditions like nonalcoholic steatohepatitis (NASH), liver fibrosis, cirrhosis, and hepatocellular carcinoma. Diagnosing and staging NAFLD presents challenges due to its non-specific symptoms and the invasive nature of liver biopsies. Our research introduces a novel artificial intelligence cascade model employing ensemble learning and feature fusion techniques. We developed a non-invasive, robust, and reliable diagnostic artificial intelligence tool that utilizes anthropometric and laboratory parameters, facilitating early detection and intervention in NAFLD progression. Our novel artificial intelligence achieved an 86% accuracy rate for the NASH steatosis staging task (non-NASH, steatosis grade 1, steatosis grade 2, and steatosis grade 3) and an impressive 96% AUC-ROC for distinguishing between NASH (steatosis grade 1, grade 2, and grade3) and non-NASH cases, outperforming current state-of-the-art models. This notable improvement in diagnostic performance underscores the potential application of artificial intelligence in the early diagnosis and treatment of NAFLD, leading to better patient outcomes and a reduced healthcare burden associated with advanced liver disease.

The quest to understand gravity's role in shaping the universe has led to the exploration of modified gravity theories. One such theory is Myrzakulov gravity, which incorporates both curvature and torsion. In this work, we investigate the effects of torsion within the framework of $f(R,T)$-gravity, a modification of General Relativity that includes both curvature and torsion. We present this theory in the Vielbein formalism, which offers a more flexible, geometric perspective on gravity. This formalism is particularly useful in Weitzenb"ock spacetime, where torsion significantly influences gravitational interactions. Our study aims to extend Myrzakulov's theory and explore its cosmological and astrophysical implications. We examine the effects of torsion on phenomena like black holes, gravitational waves, and neutron stars. These modifications could lead to observable deviations in black hole thermodynamics, gravitational wave propagation, and dense matter structure. This work provides new insights into the nature of spacetime and gravity, offering a fresh perspective on the fundamental forces governing the cosmos. It paves the way for future studies, both observational and theoretical, that could reveal new physics beyond General Relativity., Comment: 44 pages, shorter abstract

We present DiffVox, a self-supervised framework for Cone-Beam Computed Tomography (CBCT) reconstruction by directly optimizing a voxelgrid representation using physics-based differentiable X-ray rendering. Further, we investigate how the different implementations of the X-ray image formation model in the renderer affect the quality of 3D reconstruction and novel view synthesis. When combined with our regularized voxel-based learning framework, we find that using an exact implementation of the discrete Beer-Lambert law for X-ray attenuation in the renderer outperforms both widely used iterative CBCT reconstruction algorithms and modern neural field approaches, particularly when given only a few input views. As a result, we reconstruct high-fidelity 3D CBCT volumes from fewer X-rays, potentially reducing ionizing radiation exposure and improving diagnostic utility. Our implementation is available at https://github.com/hossein-momeni/DiffVox.

Machine learning models are advancing circuit design, particularly in analog circuits. They typically generate netlists that lack human interpretability. This is a problem as human designers heavily rely on the interpretability of circuit diagrams or schematics to intuitively understand, troubleshoot, and develop designs. Hence, to integrate domain knowledge effectively, it is crucial to translate ML-generated netlists into interpretable schematics quickly and accurately. We propose Schemato, a large language model (LLM) for netlist-to-schematic conversion. In particular, we consider our approach in the two settings of converting netlists to .asc files for LTSpice and LATEX files for CircuiTikz schematics. Experiments on our circuit dataset show that Schemato achieves up to 93% compilation success rate for the netlist-to-LaTeX conversion task, surpassing the 26% rate scored by the state-of-the-art LLMs. Furthermore, our experiments show that Schemato generates schematics with a mean structural similarity index measure that is 3xhigher than the best performing LLMs, therefore closer to the reference human design.

Designing integrated circuits involves substantial complexity, posing challenges in revealing its potential applications - from custom digital cells to analog circuits. Despite extensive research over the past decades in building versatile and automated frameworks, there remains open room to explore more computationally efficient AI-based solutions. This paper introduces the graph composer GraCo, a novel method for synthesizing integrated circuits using reinforcement learning (RL). GraCo learns to construct a graph step-by-step, which is then converted into a netlist and simulated with SPICE. We demonstrate that GraCo is highly configurable, enabling the incorporation of prior design knowledge into the framework. We formalize how this prior knowledge can be utilized and, in particular, show that applying consistency checks enhances the efficiency of the sampling process. To evaluate its performance, we compare GraCo to a random baseline, which is known to perform well for smaller design space problems. We demonstrate that GraCo can discover circuits for tasks such as generating standard cells, including the inverter and the two-input NAND (NAND2) gate. Compared to a random baseline, GraCo requires 5x fewer sampling steps to design an inverter and successfully synthesizes a NAND2 gate that is 2.5x faster.

Developing efficient path integral (PI) methods for atomistic simulations of vibrational spectra in heterogeneous condensed phases and interfaces has long been a challenging task. Here, we present the h-CMD method, short for hybrid centroid molecular dynamics, that combines the recently introduced fast quasi-CMD (f-QCMD) method with fast CMD (f-CMD). In this scheme, molecules that are believed to suffer more seriously from the curvature problem of CMD, e.g., water, are treated with f-QCMD, while the rest, e.g., solid surfaces, are treated with f-CMD. To test the accuracy of the newly introduced scheme, the infrared spectra of the interfacial D2O confined in the archetypal ZIF-90 framework are simulated using h-CMD compared to a variety of other PI methods, including thermostatted ring-polymer molecular dynamics (T-RPMD) and partially adiabatic CMD as well as f-CMD and experiment as reference. Comparisons are also made to classical MD, where nuclear quantum effects are neglected entirely. Our detailed comparisons at different temperatures of 250-600 K show that h-CMD produces O-D stretches that are in close agreement with the experiment, correcting the known curvature problem and red-shifting of the stretch peaks of CMD. h-CMD also corrects the known issues associated with too artificially dampened and broadened spectra of T-RPMD, which leads to missing the characteristic doublet feature of the interfacial confined water, rendering it unsuitable for these systems. The new h-CMD method broadens the applicability of f-QCMD to heterogeneous condensed phases and interfaces, where defining curvilinear coordinates for the entire system is not feasible., Comment: 11 pages, 5 figures

In this study, we explore the dynamics of warm inflation within a non-minimally coupled Peccei-Quinn (PQ) framework and evaluate its compatibility with the de Sitter Swampland Conjecture. Our model incorporates a PQ scalar field that is non-minimally coupled to gravity, facilitating inflation through a dissipative process that sustains a thermal bath, thereby distinguishing it from conventional cold inflation. We analyze the dissipation coefficient defined as $\Gamma(T, \sigma) = C_n T^n \sigma^p M^{1-n-p}$, where $C_n$ is a dimensionless constant, $M$ is a mass scale, and $n$ and $p$ are numerical powers. Our investigation focuses on three specific cases: (a) A temperature-dependent dissipation coefficient with an inverse relation, $\Gamma = C_{-1}\,\sigma^2/T$, where $n=-1$ and $p=2$; (b) A dissipation coefficient linear in field $\phi$, $\Gamma = C_{0} \sigma$, where $n=0$ and $p=1$; and (c) A dissipation coefficient linear in temperature $T$, $\Gamma = C_{1} T$, where $n=1$ and $p=0$. By examining the slow-roll dynamics in these inflationary scenarios, we derive essential cosmological parameters, including the scalar spectral index and the tensor-to-scalar ratio. We compare our results with the latest observational data from Planck 2018. Our findings suggest that the model is consistent with observational constraints while simultaneously satisfying the de Sitter Swampland conditions., Comment: v1: 28 pages, many figures

As urban populations grow, the need for accessible urban design has become urgent. Traditional survey methods for assessing public perceptions of accessibility are often limited in scope. Crowdsourcing via online reviews offers a valuable alternative to understanding public perceptions, and advancements in large language models can facilitate their use. This study uses Google Maps reviews across the United States and fine-tunes Llama 3 model with the Low-Rank Adaptation technique to analyze public sentiment on accessibility. At the POI level, most categories -- restaurants, retail, hotels, and healthcare -- show negative sentiments. Socio-spatial analysis reveals that areas with higher proportions of white residents and greater socioeconomic status report more positive sentiment, while areas with more elderly, highly-educated residents exhibit more negative sentiment. Interestingly, no clear link is found between the presence of disabilities and public sentiments. Overall, this study highlights the potential of crowdsourcing for identifying accessibility challenges and providing insights for urban planners.

Selective adsorptive capture and separation of chemically inert Kr and Xe noble gases with very low ppmv concentrations in air and industrial off-gases constitute an important technological challenge. Here, using a synergistic combination of experiment and theory, the microporous crystalline vanadomolybdates (MoVOx) as highly selective Kr sorbents are studied in detail. By varying the Mo/V ratios, we show for the first time that their one-dimensional pores can be fine-tuned for the size-selective adsorption of Kr over the larger Xe with selectivities reaching >100. Using extensive electronic structure calculations and grand canonical Monte-Carlo simulations, the competition between Kr uptake with CO2 and N2 was also investigated. As most materials reported so far are selective toward the larger, more polarizable Xe than Kr, this work constitutes an important step toward robust Kr-selective sorbent materials. This work highlights the potential use of porous crystalline transition metal oxides as energy-efficient and selective noble gas capture sorbents for industrial applications.

Using a representative translational dimer model, high throughput calculations are implemented for fast screening of a total of 74 diacenaphtho-extended heterocycle (DAH) derivatives as hole transport layer (HTL) materials in perovskite solar cells (PVSCs). Different electronic properties, including band structures, band gaps, and band edges compared to methylammonium and formamidinium lead iodide perovskites, along with reorganization energies, electronic couplings, and hole mobilities are calculated in order to decipher the effects of different parameters, including the polarity, steric and pi-conjugation, as well as the presence of explicit hydrogen bond interactions on the computed carrier mobilities of the studied materials. Full crystal structure predictions and hole mobility calculations of the top candidates resulted in some mobilities exceeding 10 cm2/V.s, further validating the employed translational dimer model as a robust approach for inverse design and fast high throughput screening of new HTL organic semiconductors with superior properties. The studied models and simulations performed in this work are instructive in designing next-generation HTL materials for higher-performance PVSCs.

School districts historically approached conflict-resolution from a zero-sum perspective: suspend students seen as disruptive and potentially harm them, or avoid suspensions and harm their classmates. Restorative practices (RP) -- focused on reparation and shared ownership of disciplinary justice -- are designed to avoid this trade-off by addressing undesirable behavior without imparting harm. This study examines Chicago Public Schools' adoption of RP. We identify decreased suspensions, improved school climate, and find no evidence of increased classroom disruption. We estimate a 19% decrease in arrests, including for violent offenses, with reduced arrests outside of school, providing evidence that RP substantively changed behavior. [Additional funding provided by the Becker Friedman Institute for Economics at the University of Chicago and the Hymen Milgrom Supporting Organization through the Successful Pathways from School to Work initiative of the University of Chicago.]

Physical neural networks (PNNs) are a class of neural-like networks that leverage the properties of physical systems to perform computation. While PNNs are so far a niche research area with small-scale laboratory demonstrations, they are arguably one of the most underappreciated important opportunities in modern AI. Could we train AI models 1000x larger than current ones? Could we do this and also have them perform inference locally and privately on edge devices, such as smartphones or sensors? Research over the past few years has shown that the answer to all these questions is likely "yes, with enough research": PNNs could one day radically change what is possible and practical for AI systems. To do this will however require rethinking both how AI models work, and how they are trained - primarily by considering the problems through the constraints of the underlying hardware physics. To train PNNs at large scale, many methods including backpropagation-based and backpropagation-free approaches are now being explored. These methods have various trade-offs, and so far no method has been shown to scale to the same scale and performance as the backpropagation algorithm widely used in deep learning today. However, this is rapidly changing, and a diverse ecosystem of training techniques provides clues for how PNNs may one day be utilized to create both more efficient realizations of current-scale AI models, and to enable unprecedented-scale models., Comment: 29 pages, 4 figures

In this work, our goals are two fold: large-vocabulary continuous sign language recognition (CSLR), and sign language retrieval. To this end, we introduce a multi-task Transformer model, CSLR2, that is able to ingest a signing sequence and output in a joint embedding space between signed language and spoken language text. To enable CSLR evaluation in the large-vocabulary setting, we introduce new dataset annotations that have been manually collected. These provide continuous sign-level annotations for six hours of test videos, and will be made publicly available. We demonstrate that by a careful choice of loss functions, training the model for both the CSLR and retrieval tasks is mutually beneficial in terms of performance -- retrieval improves CSLR performance by providing context, while CSLR improves retrieval with more fine-grained supervision. We further show the benefits of leveraging weak and noisy supervision from large-vocabulary datasets such as BOBSL, namely sign-level pseudo-labels, and English subtitles. Our model significantly outperforms the previous state of the art on both tasks.

In this work, we study several extensions of the higher curvature modification of $R^{2}$ inflation in the context of gravity's rainbow. We modify the $(R+R^{2})$ model by adding an $f_{1}R^3$-term, an $f_{2}R^4$-term, and an $f_{3}R^{3/2}$-term to the original model. We calculate the inflationary observables and confront them using the latest observational bounds from Planck 2018 data. We assume the rainbow function of the form $\tilde{f}=1+\left(\frac{H}{M}\right)^{\lambda }$ with $\lambda$ being a rainbow parameter and $M$ a mass-dimensional parameter. We demonstrate that the power spectrum of curvature perturbation relies on the dimensionless coefficient $f_{i},\,i=1,2,3$, a rainbow parameter $\lambda$ and a ratio $H/M$. Likewise, the scalar spectral index $n_s$ is affected by both $f_{i}$ and the rainbow parameter. Moreover, the tensor-to-scalar ratio $r$ is solely determined by the rainbow parameter. Interestingly, by ensuring that $n_s$ aligns with the Planck collaboration's findings at the $1\sigma$ confidence level, the tensor-to-scalar ratio could reach up to $r\sim 0.01$, which is possibly measurable for detection in forthcoming Stage IV CMB ground experiments and is certainly feasible for future dedicated space missions., Comment: v3: 21 pages, 4 figures. references added, texts modified, results unchanged. Version published in EPJC. arXiv admin note: text overlap with arXiv:2404.14766

Inspired by recent developments in the study of the model of double scaled SYK (DSSYK), as elucidated in a recent paper, we embark on a re-evaluation of the Sachdev-Ye-Kitaev (SYK) model. Our motivation stems from the insights gained from the DSSYK model, particularly its ability to capture essential features of quantum dynamics and gravitational effects. In this work, we delve into the SYK model, uncovering precise solutions for the two-point function and self-energy that have not been previously reported. Building upon the advancements made in particle physics phenomenology, we extend the SYK model to encompass tensor field theory. Through the incorporation of a cutoff term to ensure convergence, we substantially advance our understanding of quantum many-body physics. Our investigation extends to experimental parameter estimation and the exploration of cutoff dependency in random couplings, providing invaluable insights into system dynamics. The introduction of a novel tensor field theory replaces conventional fermionic degrees of freedom with tensorial counterparts, leading to the discovery of intriguing phase transition phenomena characterized by a first-order transition. Furthermore, we elucidate a direct linear relationship between the coupling parameter and the cutoff scale. These findings not only shed light on emergent behavior across both high-energy physics and condensed matter systems but also pave the way for further theoretical and experimental exploration, inspired by the recent advancements in the SYK model., Comment: 43 pages, two References added

Recent advances in the field of condensed-matter physics have unlocked the potential to realize and control emergent material phases that do not exist in thermal equilibrium. One of the most promising concepts in this regard is Floquet engineering, the coherent dressing of matter via time-periodic perturbations. However, the broad applicability of Floquet engineering to quantum materials is still unclear. For the paradigmatic case of monolayer graphene, the theoretically predicted Floquet-induced effects, despite a seminal report of the light-induced anomalous Hall effect, have been put into question. Here, we overcome this problem by using electronic structure measurements to provide direct experimental evidence of Floquet engineering in graphene. We report light-matter-dressed Dirac bands by measuring the contribution of Floquet sidebands, Volkov sidebands, and their quantum path interference to graphene's photoemission spectral function. Our results finally demonstrate that Floquet engineering in graphene is possible, paving the way for the experimental realization of the many theoretical proposals on Floquet-engineered band structures and topological phases., Comment: 4 main figures, 3 extended figures

The $d$-Cut problem is to decide if a graph has an edge cut such that each vertex has at most $d$ neighbours at the opposite side of the cut. If $d=1$, we obtain the intensively studied Matching Cut problem. The $d$-Cut problem has been studied as well, but a systematic study for special graph classes was lacking. We initiate such a study and consider classes of bounded diameter, bounded radius and $H$-free graphs. We prove that for all $d\geq 2$, $d$-Cut is polynomial-time solvable for graphs of diameter $2$, $(P_3+P_4)$-free graphs and $P_5$-free graphs. These results extend known results for $d=1$. However, we also prove several NP-hardness results for $d$-Cut that contrast known polynomial-time results for $d=1$. Our results lead to full dichotomies for bounded diameter and bounded radius and to almost-complete dichotomies for $H$-free graphs.

We investigate static cylindrical solutions within an extended theory of modified gravity. By incorporating various coupling functions through a straightforward boost symmetry approach, we establish the equations of motion in a self-consistent manner and subsequently determine the linear scalar field profile. Utilizing analytical methods, we solve the system of equations for the metric functions and the $U(1)$ gauge field, revealing their dependence on Bessel's functions. To comprehend gravito-objects exhibiting cylindrical symmetry, we develop a perturbative framework aimed at identifying all nontrivial solutions for the scalar profiles. Introducing first-order truncated perturbation equations for the gauge field, synchronized with metric gauges and electromagnetic field considerations, we demonstrate their integrability and obtain solutions through quadrature. Our findings suggest the feasibility of obtaining self-gravitating cylindrical structures within the scalar-vector-tensor theory., Comment: v3: version published in Nucl. Phys. B

Blind System Identification (BSI) is used to extract a system model whenever input data is not attainable. Therefore, the input data and system model should be estimated simultaneously. Because of nonlinearities in a large number of systems, BSI problem is usually challenging to solve. In this paper, an innovative solution is proposed to deal with the BSI problem in nonlinear systems using the properties of the Linear Parameter-Varying (LPV) systems and Hidden Markov Models (HMM). More specifically, assuming scheduling variable is not measurable, the dynamic of the LPV system is approximated. To solve the BSI problem in this context, LPV structure is modeled as an HMM network and a modified Quasi-Static combination of Viterbi and Baum-Welch algorithms (QSVBW) is proposed to estimate the nonlinear mappings and scheduling variable signal. The applicability and the performance of the suggested QSVBW algorithm has been justified by numerical studies.

The shadows and photon spheres are investigated for both static and rotating traversable wormholes. In particular, with a red-shift function constructed in combination with two distinct shape functions, the null geodesic equation is found to predict the light trajectory. To explore observable phenomena and gain a deeper understanding of the gravitational influences originating from wormholes, we explicitly demonstrate the shadow produced and photon spheres associated with both static and rotating traversable wormholes by taking into account the gravitational lensing effects and photon motion in strong gravitational fields. Specifically, the ray tracing of the light geodesic and profiles of the intensity around both static and rotating traversable wormholes can be accomplished., Comment: V1: 18 pages, 6 captioned figures

Keystroke biometrics is a promising approach for user identification and verification, leveraging the unique patterns in individuals' typing behavior. In this paper, we propose a Transformer-based network that employs self-attention to extract informative features from keystroke sequences, surpassing the performance of traditional Recurrent Neural Networks. We explore two distinct architectures, namely bi-encoder and cross-encoder, and compare their effectiveness in keystroke authentication. Furthermore, we investigate different loss functions, including triplet, batch-all triplet, and WDCL loss, along with various distance metrics such as Euclidean, Manhattan, and cosine distances. These experiments allow us to optimize the training process and enhance the performance of our model. To evaluate our proposed model, we employ the Aalto desktop keystroke dataset. The results demonstrate that the bi-encoder architecture with batch-all triplet loss and cosine distance achieves the best performance, yielding an exceptional Equal Error Rate of 0.0186%. Furthermore, alternative algorithms for calculating similarity scores are explored to enhance accuracy. Notably, the utilization of a one-class Support Vector Machine reduces the Equal Error Rate to an impressive 0.0163%. The outcomes of this study indicate that our model surpasses the previous state-of-the-art in free-text keystroke authentication. These findings contribute to advancing the field of keystroke authentication and offer practical implications for secure user verification systems.

We report laboratory experiments and numerical simulations demonstrating that the anisotropic characteristics of rocks play a major role in the elongation of hydraulic fractures propagating in a plane perpendicular to bedding. Transverse anisotropy leads to larger hydraulic fracture extension in the parallel-to-bedding/divider direction compared to the perpendicular-to-bedding/arrester direction. This directly promotes vertical containment of hydraulic fractures in most sedimentary basins worldwide even in the absence of any favorable in-situ stress contrasts or other material heterogeneities. More importantly, the ratio of the energy dissipated in fluid viscous flow in the fracture to the energy dissipated in the creation of new surfaces is found to play a critical role on fracture elongation, with fracture-energy dominated hydraulic fractures being the most elongated while the viscous dominated ones remain more circular. These results open the door to a better engineering and control of hydraulic fractures containment at depth in view of the competition between material anisotropy and injection parameters (fluid viscosity and rate of injection).