Search

Multimodal AI has the potential to significantly enhance document-understanding tasks, such as processing receipts, understanding workflows, extracting data from documents, and summarizing reports. Code generation tasks that require long-structured outputs can also be enhanced by multimodality. Despite this, their use in commercial applications is often limited due to limited access to training data and restrictive licensing, which hinders open access. To address these limitations, we introduce BigDocs-7.5M, a high-quality, open-access dataset comprising 7.5 million multimodal documents across 30 tasks. We use an efficient data curation process to ensure our data is high-quality and license-permissive. Our process emphasizes accountability, responsibility, and transparency through filtering rules, traceable metadata, and careful content analysis. Additionally, we introduce BigDocs-Bench, a benchmark suite with 10 novel tasks where we create datasets that reflect real-world use cases involving reasoning over Graphical User Interfaces (GUI) and code generation from images. Our experiments show that training with BigDocs-Bench improves average performance up to 25.8% over closed-source GPT-4o in document reasoning and structured output tasks such as Screenshot2HTML or Image2Latex generation. Finally, human evaluations showed a preference for outputs from models trained on BigDocs over GPT-4o. This suggests that BigDocs can help both academics and the open-source community utilize and improve AI tools to enhance multimodal capabilities and document reasoning. The project is hosted at https://bigdocs.github.io ., Comment: The project is hosted at https://bigdocs.github.io

This paper proposes the joint use of digital self-interference cancellation (DSIC) and spatial suppression to mitigate far-field self-interference (SI) in full-duplex multiple-input multiple-output (MIMO) systems. Far-field SI, caused by echoes from environmental scatterers, is modeled based on the scatterers' angle and delay parameters, stored in a scatterer map. For each scatterer, the most suitable action regarding communication is selected from transmit beamforming, receive beamforming, DSIC, and no-action. This selection is based on simple metrics that show the expected uplink and downlink communication performance. In addition, emerging scatterers that deteriorate the communication are detected, and their delay and angles are acquired, providing an up-to-date scatterer map and presenting a \emph{sensing for communication} case. The proposed selection policy is compared with the individual implementations of DSIC and spatial suppression, highlighting the failure cases for each. It is shown that the proposed policy stays unaffected in these problematic cases and achieves SI-free performance., Comment: Partially presented in IEEE PIMRC in September 2024. To be submitted to IEEE Transactions on Communications

Distributed quantum metrology (DQM) enables the estimation of global functions of d distributed parameters beyond the capability of separable sensors. To estimate an arbitrary analytic function of the parameters it suffices that a DQM scheme can measure an arbitrary linear combination of these parameters, and a number of schemes have now been devised to achieve this at the Heisenberg limit. The most practical to-date requires d coherent inputs and one squeezed vacuum. Here we provide a full understanding of the minimal input resources and linear networks required to achieve DQM of arbitrary functions. We are able to fully elucidate the structure of any linear network for DQM that has two non-vacuum inputs, and we show that two non-vacuum inputs, one non-classical, is the minimum required to achieve DQM with arbitrary weights at the Heisenberg limit. We characterize completely the properties of the nonclassical input required to obtain a quantum advantage, showing that a wide range of inputs make this possible, including a squeezed vacuum. We further elucidate two distinct regimes of the distributed sensing network. The first achieves Heisenberg scaling. In the second the nonclassical input is much weaker than the coherent input, nevertheless providing a significant quantum enhancement to the otherwise classical sensitivity., Comment: 10 pages including supplemental materials, 3 figures

We introduce a diffusion model for Gaussian Splats, SplatDiffusion, to enable generation of three-dimensional structures from single images, addressing the ill-posed nature of lifting 2D inputs to 3D. Existing methods rely on deterministic, feed-forward predictions, which limit their ability to handle the inherent ambiguity of 3D inference from 2D data. Diffusion models have recently shown promise as powerful generative models for 3D data, including Gaussian splats; however, standard diffusion frameworks typically require the target signal and denoised signal to be in the same modality, which is challenging given the scarcity of 3D data. To overcome this, we propose a novel training strategy that decouples the denoised modality from the supervision modality. By using a deterministic model as a noisy teacher to create the noised signal and transitioning from single-step to multi-step denoising supervised by an image rendering loss, our approach significantly enhances performance compared to the deterministic teacher. Additionally, our method is flexible, as it can learn from various 3D Gaussian Splat (3DGS) teachers with minimal adaptation; we demonstrate this by surpassing the performance of two different deterministic models as teachers, highlighting the potential generalizability of our framework. Our approach further incorporates a guidance mechanism to aggregate information from multiple views, enhancing reconstruction quality when more than one view is available. Experimental results on object-level and scene-level datasets demonstrate the effectiveness of our framework.

The response of many-body quantum systems to an optical pulse can be extremely challenging to model. Here we explore the use of neural networks, both traditional and generative, to learn and thus simulate the response of such a system from data. The quantum system can be viewed as performing a complex mapping from an input time-series (the optical pulse) to an output time-series (the systems response) which is often also an optical pulse. Using both the transverse and non-integrable Ising models as examples, we show that not only can temporal convolutional networks capture the input/output mapping generated by the system but can also be used to characterize the complexity of the mapping. This measure of complexity is provided by the size of the smallest latent space that is able to accurately model the mapping. We further find that a generative model, in particular a variational auto-encoder, significantly outperforms traditional auto-encoders at learning the complex response of many-body quantum systems. For the example that generated the most complex mapping, the variational auto-encoder produces outputs that have less than 10% error for more than 90% of inputs across our test data.

Electron microscopy techniques such as electron energy-loss spectroscopy (EELS) facilitate the spatio-spectral characterization of plasmonic nanostructures. In this work, a time-dependent perspective is presented, which significantly enhances the utility of EELS. Specifically, silver nanowires offer the material and geometric features for various high-quality plasmonic excitations. This provides an ideal illustrative system for combined experimental-theoretical analyses of the different plasmonic excitations and their real-time dynamics. It is demonstrated how the plasmonic excitations propagating inside the wire repeatedly interact with the swift electrons in an EELS configuration. In addition, the role of azimuthal modes, often overlooked for very thin wires, is observed and analyzed in both the energy-loss spectrum and the dynamical perspective. Such a complete understanding of the interaction of electrons and plasmonic excitation is key for the design of efficient plasmonic sensors, the study of hot electron dynamics in metals, and applications in the context of electron quantum optics, where full control of the spatial and temporal characteristics of the fields at the nanometer and femtosecond scales is highly desirable.

The integration of machine learning in magnetic resonance imaging (MRI), specifically in neuroimaging, is proving to be incredibly effective, leading to better diagnostic accuracy, accelerated image analysis, and data-driven insights, which can potentially transform patient care. Deep learning models utilize multiple layers of processing to capture intricate details of complex data, which can then be used on a variety of tasks, including brain tumor classification, segmentation, image synthesis, and registration. Previous research demonstrates high accuracy in tumor segmentation using various model architectures, including nn-UNet and Swin-UNet. U-Mamba, which uses state space modeling, also achieves high accuracy in medical image segmentation. To leverage these models, we propose a deep learning framework that ensembles these state-of-the-art architectures to achieve accurate segmentation and produce finely synthesized images.

Scientists and engineers use simulators to model empirically observed phenomena. However, tuning the parameters of a simulator to ensure its outputs match observed data presents a significant challenge. Simulation-based inference (SBI) addresses this by enabling Bayesian inference for simulators, identifying parameters that match observed data and align with prior knowledge. Unlike traditional Bayesian inference, SBI only needs access to simulations from the model and does not require evaluations of the likelihood-function. In addition, SBI algorithms do not require gradients through the simulator, allow for massive parallelization of simulations, and can perform inference for different observations without further simulations or training, thereby amortizing inference. Over the past years, we have developed, maintained, and extended $\texttt{sbi}$, a PyTorch-based package that implements Bayesian SBI algorithms based on neural networks. The $\texttt{sbi}$ toolkit implements a wide range of inference methods, neural network architectures, sampling methods, and diagnostic tools. In addition, it provides well-tested default settings but also offers flexibility to fully customize every step of the simulation-based inference workflow. Taken together, the $\texttt{sbi}$ toolkit enables scientists and engineers to apply state-of-the-art SBI methods to black-box simulators, opening up new possibilities for aligning simulations with empirically observed data.

This paper introduces a scenario where a maneuverable satellite in geostationary orbit (GEO) conducts on-orbit attacks, targeting communication between a GEO satellite and a ground station, with the ability to switch between stationary and time-variant jamming modes. We propose a machine learning-based detection approach, employing the random forest algorithm with principal component analysis (PCA) to enhance detection accuracy in the stationary model. At the same time, an adaptive threshold-based technique is implemented for the time-variant model to detect dynamic jamming events effectively. Our methodology emphasizes the need for the use of orbital dynamics in integrating physical constraints from satellite dynamics to improve model robustness and detection accuracy. Simulation results highlight the effectiveness of PCA in enhancing the performance of the stationary model, while the adaptive thresholding method achieves high accuracy in detecting jamming in the time-variant scenario. This approach provides a robust solution for mitigating the evolving threats to satellite communication in GEO environments.

Infrared (IR) imaging offers advantages in several fields due to its unique ability of capturing content in extreme light conditions. However, the demanding hardware requirements of high-resolution IR sensors limit its widespread application. As an alternative, visible light can be used to synthesize IR images but this causes a loss of fidelity in image details and introduces inconsistencies due to lack of contextual awareness of the scene. This stems from a combination of using visible light with a standard dynamic range, especially under extreme lighting, and a lack of contextual awareness can result in pseudo-thermal-crossover artifacts. This occurs when multiple objects with similar temperatures appear indistinguishable in the training data, further exacerbating the loss of fidelity. To solve this challenge, this paper proposes CapHDR2IR, a novel framework incorporating vision-language models using high dynamic range (HDR) images as inputs to generate IR images. HDR images capture a wider range of luminance variations, ensuring reliable IR image generation in different light conditions. Additionally, a dense caption branch integrates semantic understanding, resulting in more meaningful and discernible IR outputs. Extensive experiments on the HDRT dataset show that the proposed CapHDR2IR achieves state-of-the-art performance compared with existing general domain transfer methods and those tailored for visible-to-infrared image translation.

Exposure correction methods aim to adjust the luminance while maintaining other luminance-unrelated information. However, current exposure correction methods have difficulty in fully separating luminance-related and luminance-unrelated components, leading to distortions in color, loss of detail, and requiring extra restoration procedures. Inspired by principal component analysis (PCA), this paper proposes an exposure correction method called luminance component analysis (LCA). LCA applies the orthogonal constraint to a U-Net structure to decouple luminance-related and luminance-unrelated features. With decoupled luminance-related features, LCA adjusts only the luminance-related components while keeping the luminance-unrelated components unchanged. To optimize the orthogonal constraint problem, LCA employs a geometric optimization algorithm, which converts the constrained problem in Euclidean space to an unconstrained problem in orthogonal Stiefel manifolds. Extensive experiments show that LCA can decouple the luminance feature from the RGB color space. Moreover, LCA achieves the best PSNR (21.33) and SSIM (0.88) in the exposure correction dataset with 28.72 FPS.

This paper deals with the long-term behavior of positive solutions for the following parabolic-elliptic chemotaxis competition system with weak singular sensitivity and logistic source \begin{equation} \label{abstract-eq} \begin{cases} u_t=\Delta u-\chi \nabla\cdot (\frac{u}{v^{\lambda}} \nabla v) +ru- \mu u^2, \quad &x\in \Omega,\cr 0=\Delta v- \alpha v +\beta u,\quad &x\in \Omega, \cr \frac{\partial u}{\partial \nu}=\frac{\partial v}{\partial \nu}=0,\quad &x\in\partial\Omega, \end{cases}\, \end{equation} where $\Omega \subset \mathbb{R}^N (N \ge 2)$ is a smooth bounded domain, the parameters $\chi,\, r, \, \mu, \, \alpha,\,\beta$ are positive constants and $\lambda \in (0,1).$ In this article, for all suitably smooth initial data $u_0\in C^0(\bar\Omega)$ with $u_0 \not \equiv 0,$ it has been proven that: First, there exists $\mu > \mu_1^*(p,\lambda,\chi,\beta)$ such that any globally defined positive solution is $L^p(\Omega)$-bounded with $p \ge 2.$ Next, there exists $\mu > \mu_2^*(N,\lambda,\chi,\beta)$ such that any globally defined classical solutions is globally bounded. Third, there exists $\mu > \mu_3^*(N,\lambda,\chi,\beta)$ such that any globally defined positive solution is uniformly bounded above and below eventually by some positive constants that are independent of its initial function $u_0.$ Last, there exists $\mu > \mu_4^*(N,\lambda,\chi,\alpha,\beta,r,\Omega)$ such that any globally bounded classical solution to system (0.1) exponentially converges to the constant steady state $(\frac{r}{\mu},\frac{\beta}{\alpha}\frac{r}{\mu}).$

The rapid evolution of communication technologies, compounded by recent geopolitical events such as the Viasat cyberattack in February 2022, has highlighted the urgent need for fast and reliable satellite missions for military and civil security operations. Consequently, this paper examines two Earth observation (EO) missions: one utilizing a single low Earth orbit (LEO) satellite and another through a network of LEO satellites, employing a secure-by-component design strategy. This approach begins by defining the scope of technical security engineering, decomposing the system into components and data flows, and enumerating attack surfaces. Then it proceeds by identifying threats to low-level components, applying secure-by-design principles, redesigning components into secure blocks in alignment with the Space Attack Research & Tactic Analysis (SPARTA) framework, and crafting shall statements to refactor the system design, with a particular focus on improving the security of the link segment.

We present DeSiRe-GS, a self-supervised gaussian splatting representation, enabling effective static-dynamic decomposition and high-fidelity surface reconstruction in complex driving scenarios. Our approach employs a two-stage optimization pipeline of dynamic street Gaussians. In the first stage, we extract 2D motion masks based on the observation that 3D Gaussian Splatting inherently can reconstruct only the static regions in dynamic environments. These extracted 2D motion priors are then mapped into the Gaussian space in a differentiable manner, leveraging an efficient formulation of dynamic Gaussians in the second stage. Combined with the introduced geometric regularizations, our method are able to address the over-fitting issues caused by data sparsity in autonomous driving, reconstructing physically plausible Gaussians that align with object surfaces rather than floating in air. Furthermore, we introduce temporal cross-view consistency to ensure coherence across time and viewpoints, resulting in high-quality surface reconstruction. Comprehensive experiments demonstrate the efficiency and effectiveness of DeSiRe-GS, surpassing prior self-supervised arts and achieving accuracy comparable to methods relying on external 3D bounding box annotations. Code is available at \url{https://github.com/chengweialan/DeSiRe-GS}

Accurate and fast recognition of forgeries is an issue of great importance in the fields of artificial intelligence, image processing and object detection. Recognition of forgeries of facial imagery is the process of classifying and defining the faces in it by analyzing real-world facial images. This process is usually accomplished by extracting features from an image, using classifier algorithms, and correctly interpreting the results. Recognizing forgeries of facial imagery correctly can encounter many different challenges. For example, factors such as changing lighting conditions, viewing faces from different angles can affect recognition performance, and background complexity and perspective changes in facial images can make accurate recognition difficult. Despite these difficulties, significant progress has been made in the field of forgery detection. Deep learning algorithms, especially Convolutional Neural Networks (CNNs), have significantly improved forgery detection performance. This study focuses on image processing-based forgery detection using Fake-Vs-Real-Faces (Hard) [10] and 140k Real and Fake Faces [61] data sets. Both data sets consist of two classes containing real and fake facial images. In our study, two lightweight deep learning models are proposed to conduct forgery detection using these images. Additionally, 8 different pretrained CNN architectures were tested on both data sets and the results were compared with newly developed lightweight CNN models. It's shown that the proposed lightweight deep learning models have minimum number of layers. It's also shown that the proposed lightweight deep learning models detect forgeries of facial imagery accurately, and computationally efficiently. Although the data set consists only of face images, the developed models can also be used in other two-class object recognition problems., Comment: 13 pages, 6 figures, 10 tables

The increasing demand for data usage in wireless communications requires using wider bands in the spectrum, especially for backhaul links. Yet, allocations in the spectrum for non-communication systems inhibit merging bands to achieve wider bandwidth. To overcome this issue, spectrum-sharing or opportunistic spectrum utilization by secondary users stands out as a promising solution. However, both approaches must minimize interference to primary users. Therefore, spectrum sensing becomes vital for such opportunistic usage, ensuring the proper operation of the primary users. Although this problem has been investigated for 2D networks, unmanned aerial vehicle (UAV) networks need different points of view concerning 3D space, its challenges, and opportunities. For this purpose, we propose a federated learning (FL)-based method for spectrum sensing in UAV networks to account for their distributed nature and limited computational capacity. FL enables local training without sharing raw data while guaranteeing the privacy of local users,lowering communication overhead, and increasing data diversity. Furthermore, we develop a federated aggregation method, namely FedSNR, that considers the signal-to-noise ratio observed by UAVs to acquire a global model. The numerical results show that the proposed architecture and the aggregation method outperform traditional methods.

Some research now suggests that ChatGPT can estimate the quality of journal articles from their titles and abstracts. This has created the possibility to use ChatGPT quality scores, perhaps alongside citation-based formulae, to support peer review for research evaluation. Nevertheless, ChatGPT's internal processes are effectively opaque, despite it writing a report to support its scores, and its biases are unknown. This article investigates whether publication date and field are biasing factors. Based on submitting a monodisciplinary journal-balanced set of 117,650 articles from 26 fields published in the years 2003, 2008, 2013, 2018 and 2023 to ChatGPT 4o-mini, the results show that average scores increased over time, and this was not due to author nationality or title and abstract length changes. The results also varied substantially between fields, and first author countries. In addition, articles with longer abstracts tended to receive higher scores, but plausibly due to such articles tending to be better rather than due to ChatGPT analysing more text. Thus, for the most accurate research quality evaluation results from ChatGPT, it is important to normalise ChatGPT scores for field and year and check for anomalies caused by sets of articles with short abstracts.

Emerging Large Language Model (LLM) applications require long input prompts to perform complex downstream tasks like document analysis and code generation. For these long context length applications, the length of the input prompt poses a significant challenge in terms of inference efficiency since the inference costs increase linearly with sequence length. However, for many of these applications, much of the context in the prompt is fixed across different user inputs, thereby providing the opportunity to perform offline optimizations to process user inputs quickly, as they are received. In this work, we propose Squeezed Attention as a mechanism to accelerate LLM applications where a large portion of the input prompt is fixed. We first leverage K-means clustering offline to group the keys for the fixed context based on semantic similarity and represent each cluster with a single centroid value. During inference, we compare query tokens from the user input with the centroids to predict which of the keys from the fixed context are semantically relevant and need to be loaded during inference. We then compute exact attention using only these important keys from the fixed context, thereby reducing bandwidth and computational costs. We also extend our method to use a hierarchical centroid lookup to identify important keys, which can reduce the complexity of attention from linear to logarithmic with respect to the context length. We implement optimized Triton kernels for centroid comparison and sparse FlashAttention with important keys, achieving more than 4x speedups during both the prefill and generation phases for long-context inference. Furthermore, we have extensively evaluated our method on various long-context benchmarks including LongBench, where it achieves a 3x reduction in KV cache budget without accuracy loss and up to an 8x reduction with <0.5 point accuracy gap for various models.

White matter alterations are increasingly implicated in neurological diseases and their progression. International-scale studies use diffusion-weighted magnetic resonance imaging (DW-MRI) to qualitatively identify changes in white matter microstructure and connectivity. Yet, quantitative analysis of DW-MRI data is hindered by inconsistencies stemming from varying acquisition protocols. There is a pressing need to harmonize the preprocessing of DW-MRI datasets to ensure the derivation of robust quantitative diffusion metrics across acquisitions. In the MICCAI-CDMRI 2023 QuantConn challenge, participants were provided raw data from the same individuals collected on the same scanner but with two different acquisitions and tasked with preprocessing the DW-MRI to minimize acquisition differences while retaining biological variation. Submissions are evaluated on the reproducibility and comparability of cross-acquisition bundle-wise microstructure measures, bundle shape features, and connectomics. The key innovations of the QuantConn challenge are that (1) we assess bundles and tractography in the context of harmonization for the first time, (2) we assess connectomics in the context of harmonization for the first time, and (3) we have 10x additional subjects over prior harmonization challenge, MUSHAC and 100x over SuperMUDI. We find that bundle surface area, fractional anisotropy, connectome assortativity, betweenness centrality, edge count, modularity, nodal strength, and participation coefficient measures are most biased by acquisition and that machine learning voxel-wise correction, RISH mapping, and NeSH methods effectively reduce these biases. In addition, microstructure measures AD, MD, RD, bundle length, connectome density, efficiency, and path length are least biased by these acquisition differences., Comment: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024/019

Genealogy, the study of family history and lineage, has seen tremendous growth over the past decade, fueled by technological advances such as home DNA testing and mass digitization of historical records. However, HCI research on genealogy practices is nascent, with the most recent major studies predating this transformation. In this paper, we present a qualitative study of the current state of technological support for genealogy research, collaboration, and education. Through semi-structured interviews with 20 genealogists with diverse expertise, we report on current practices, challenges, and success stories around how genealogists conduct research, collaborate, and learn skills. We contrast the experiences of amateurs and experts, describe the emerging importance of standardization and professionalization of the field, and stress the critical role of computer systems in genealogy education. We bridge studies of sensemaking and information literacy through this empirical study on genealogy research practices, and conclude by discussing how genealogy presents a unique perspective through which to study collective sensemaking and education in online communities.

A set of $m$ indivisible goods is to be allocated to a set of $n$ agents. Each agent $i$ has an additive valuation function $v_i$ over goods. The value of a good $g$ for agent $i$ is either $1$ or $s$, where $s$ is a fixed rational number greater than one, and the value of a bundle of goods is the sum of the values of the goods in the bundle. An \emph{allocation} $X$ is a partition of the goods into bundles $X_1$, \ldots, $X_n$, one for each agent. The \emph{Nash Social Welfare} ($\NSW$) of an allocation $X$ is defined as \[ \NSW(X) = \left( \prod_i v_i(X_i) \right)^{\sfrac{1}{n}}.\] The \emph{$\NSW$-allocation} maximizes the Nash Social Welfare. In~\cite{NSW-twovalues-halfinteger} it was shown that the $\NSW$-allocation can be computed in polynomial time, if $s$ is an integer or a half-integer, and that the problem is NP-complete otherwise. The proof for the half-integer case is quite involved. In this note we give a simpler and shorter proof, Comment: arXiv admin note: text overlap with arXiv:2207.10949

Electronic health records (EHRs) are stored in various database systems with different database models on heterogeneous storage architectures, such as relational databases, document stores, or graph databases. These different database models have a big impact on query complexity and performance. While this has been a known fact in database research, its implications for the growing number of Text-to-Query systems have surprisingly not been investigated so far. In this paper, we present SM3-Text-to-Query, the first multi-model medical Text-to-Query benchmark based on synthetic patient data from Synthea, following the SNOMED-CT taxonomy -- a widely used knowledge graph ontology covering medical terminology. SM3-Text-to-Query provides data representations for relational databases (PostgreSQL), document stores (MongoDB), and graph databases (Neo4j and GraphDB (RDF)), allowing the evaluation across four popular query languages, namely SQL, MQL, Cypher, and SPARQL. We systematically and manually develop 408 template questions, which we augment to construct a benchmark of 10K diverse natural language question/query pairs for these four query languages (40K pairs overall). On our dataset, we evaluate several common in-context-learning (ICL) approaches for a set of representative closed and open-source LLMs. Our evaluation sheds light on the trade-offs between database models and query languages for different ICL strategies and LLMs. Last, SM3-Text-to-Query is easily extendable to additional query languages or real, standard-based patient databases., Comment: NeurIPS 2024 Track Datasets and Benchmarks

Query optimization has become a research area where classical algorithms are being challenged by machine learning algorithms. At the same time, recent trends in learned query optimizers have shown that it is prudent to take advantage of decades of database research and augment classical query optimizers by shrinking the plan search space through different types of hints (e.g. by specifying the join type, scan type or the order of joins) rather than completely replacing the classical query optimizer with machine learning models. It is especially relevant for cases when classical optimizers cannot fully enumerate all logical and physical plans and, as an alternative, need to rely on less robust approaches like genetic algorithms. However, even symbiotically learned query optimizers are hampered by the need for vast amounts of training data, slow plan generation during inference and unstable results across various workload conditions. In this paper, we present GenJoin - a novel learned query optimizer that considers the query optimization problem as a generative task and is capable of learning from a random set of subplan hints to produce query plans that outperform the classical optimizer. GenJoin is the first learned query optimizer that significantly and consistently outperforms PostgreSQL as well as state-of-the-art methods on two well-known real-world benchmarks across a variety of workloads using rigorous machine learning evaluations.

Touch is a crucial sensing modality that provides rich information about object properties and interactions with the physical environment. Humans and robots both benefit from using touch to perceive and interact with the surrounding environment (Johansson and Flanagan, 2009; Li et al., 2020; Calandra et al., 2017). However, no existing systems provide rich, multi-modal digital touch-sensing capabilities through a hemispherical compliant embodiment. Here, we describe several conceptual and technological innovations to improve the digitization of touch. These advances are embodied in an artificial finger-shaped sensor with advanced sensing capabilities. Significantly, this fingertip contains high-resolution sensors (~8.3 million taxels) that respond to omnidirectional touch, capture multi-modal signals, and use on-device artificial intelligence to process the data in real time. Evaluations show that the artificial fingertip can resolve spatial features as small as 7 um, sense normal and shear forces with a resolution of 1.01 mN and 1.27 mN, respectively, perceive vibrations up to 10 kHz, sense heat, and even sense odor. Furthermore, it embeds an on-device AI neural network accelerator that acts as a peripheral nervous system on a robot and mimics the reflex arc found in humans. These results demonstrate the possibility of digitizing touch with superhuman performance. The implications are profound, and we anticipate potential applications in robotics (industrial, medical, agricultural, and consumer-level), virtual reality and telepresence, prosthetics, and e-commerce. Toward digitizing touch at scale, we open-source a modular platform to facilitate future research on the nature of touch., Comment: 28 pages

Many applications of cyber-physical systems require real-time communication: manufacturing, automotive, etc. Recent Ethernet standards for Time Sensitive Networking (TSN) offer time-triggered scheduling in order to guarantee low latency and jitter bounds. This requires precise frame transmission planning, which becomes especially hard when dealing with many streams, large networks, and dynamically changing communications. A very promising approach uses conflict graphs, modeling conflicting transmission configurations. Since the creation of conflict graphs is the bottleneck in these approaches, we provide an improvement to the conflict graph creation. We present a randomized selection process that reduces the overall size of the graph in half and three heuristics to improve the scheduling success. In our evaluations we show substantial improvements in the graph creation speed and the scheduling success compared to existing work, updating existing schedules in fractions of a second. Additionally, offline planning of 9000 streams was performed successfully within minutes.

One of the major bottlenecks for efficient deployment of neural network based recommendation systems is the memory footprint of their embedding tables. Although many neural network based recommendation systems could benefit from the faster on-chip memory access and increased computational power of hardware accelerators, the large embedding tables in these models often cannot fit on the constrained memory of accelerators. Despite the pervasiveness of these models, prior methods in memory optimization and parallelism fail to address the memory and communication costs of large embedding tables on accelerators. As a result, the majority of models are trained on CPUs, while current implementations of accelerators are hindered by issues such as bottlenecks in inter-device communication and main memory lookups. In this paper, we propose a theoretical framework that analyses the communication costs of arbitrary distributed systems that use lookup tables. We use this framework to propose algorithms that maximize throughput subject to memory, computation, and communication constraints. Furthermore, we demonstrate that our method achieves strong theoretical performance across dataset distributions and memory constraints, applicable to a wide range of use cases from mobile federated learning to warehouse-scale computation. We implement our framework and algorithms in PyTorch and achieve up to 6x increases in training throughput on GPU systems over baselines, on the Criteo Terabytes dataset.

In this research, we create a scalable version of the quantum Fourier transform-based arithmetic circuit to perform addition and subtraction operations on N n-bit unsigned integers encoded in quantum registers, and it is compatible with d-level quantum sources, called qudits. We present qubit- and ququart-based multi-input QFT adders, and we compare and discuss potential benefits such as circuit simplicity and noise sensitivity. The results show that a ququart-based system significantly reduces gate count and improves computational efficiency compared to qubit-based systems. Overall, the findings presented in this study represent a promising step forward in the development of efficient quantum arithmetic circuits, particularly for multi-input operations, with clear advantages for ququart-based systems in reducing gate count, decoherence, and circuit complexity., Comment: 8 pages, 7 figures, Comments are welcome

Causal discovery with time series data remains a challenging yet increasingly important task across many scientific domains. Convergent cross mapping (CCM) and related methods have been proposed to study time series that are generated by dynamical systems, where traditional approaches like Granger causality are unreliable. However, CCM often yields inaccurate results depending upon the quality of the data. We propose the Tangent Space Causal Inference (TSCI) method for detecting causalities in dynamical systems. TSCI works by considering vector fields as explicit representations of the systems' dynamics and checks for the degree of synchronization between the learned vector fields. The TSCI approach is model-agnostic and can be used as a drop-in replacement for CCM and its generalizations. We first present a basic version of the TSCI algorithm, which is shown to be more effective than the basic CCM algorithm with very little additional computation. We additionally present augmented versions of TSCI that leverage the expressive power of latent variable models and deep learning. We validate our theory on standard systems, and we demonstrate improved causal inference performance across a number of benchmark tasks., Comment: 18 pages, 7 figures. Accepted to NeurIPS 2024

Material discovery is a critical area of research with the potential to revolutionize various fields, including carbon capture, renewable energy, and electronics. However, the immense scale of the chemical space makes it challenging to explore all possible materials experimentally. In this paper, we introduce FlowLLM, a novel generative model that combines large language models (LLMs) and Riemannian flow matching (RFM) to design novel crystalline materials. FlowLLM first fine-tunes an LLM to learn an effective base distribution of meta-stable crystals in a text representation. After converting to a graph representation, the RFM model takes samples from the LLM and iteratively refines the coordinates and lattice parameters. Our approach significantly outperforms state-of-the-art methods, increasing the generation rate of stable materials by over three times and increasing the rate for stable, unique, and novel crystals by $\sim50\%$ - a huge improvement on a difficult problem. Additionally, the crystals generated by FlowLLM are much closer to their relaxed state when compared with another leading model, significantly reducing post-hoc computational cost.

Estimated brain age from magnetic resonance image (MRI) and its deviation from chronological age can provide early insights into potential neurodegenerative diseases, supporting early detection and implementation of prevention strategies. Diffusion MRI (dMRI), a widely used modality for brain age estimation, presents an opportunity to build an earlier biomarker for neurodegenerative disease prediction because it captures subtle microstructural changes that precede more perceptible macrostructural changes. However, the coexistence of macro- and micro-structural information in dMRI raises the question of whether current dMRI-based brain age estimation models are leveraging the intended microstructural information or if they inadvertently rely on the macrostructural information. To develop a microstructure-specific brain age, we propose a method for brain age identification from dMRI that minimizes the model's use of macrostructural information by non-rigidly registering all images to a standard template. Imaging data from 13,398 participants across 12 datasets were used for the training and evaluation. We compare our brain age models, trained with and without macrostructural information minimized, with an architecturally similar T1-weighted (T1w) MRI-based brain age model and two state-of-the-art T1w MRI-based brain age models that primarily use macrostructural information. We observe difference between our dMRI-based brain age and T1w MRI-based brain age across stages of neurodegeneration, with dMRI-based brain age being older than T1w MRI-based brain age in participants transitioning from cognitively normal (CN) to mild cognitive impairment (MCI), but younger in participants already diagnosed with Alzheimer's disease (AD). Approximately 4 years before MCI diagnosis, dMRI-based brain age yields better performance than T1w MRI-based brain ages in predicting transition from CN to MCI.

We consider the contact process with dormancy, where wake-up times follow a renewal process. Without infection between dormant individuals, we show that the process under certain conditions grows at most logarithmically. On the other hand, if infections between dormant individuals are possible, the process survives with positive probability even on finite graphs., Comment: 12 pages

Large-scale recommendation models are currently the dominant workload for many large Internet companies. These recommenders are characterized by massive embedding tables that are sparsely accessed by the index for user and item features. The size of these 1TB+ tables imposes a severe memory bottleneck for the training and inference of recommendation models. In this work, we propose a novel recommendation framework that is small, powerful, and efficient to run and train, based on the state-of-the-art Deep Learning Recommendation Model (DLRM). The proposed framework makes inference more efficient on the cloud servers, explores the possibility of deploying powerful recommenders on smaller edge devices, and optimizes the workload of the communication overhead in distributed training under the data parallelism settings. Specifically, we show that quantization-aware training (QAT) can impose a strong regularization effect to mitigate the severe overfitting issues suffered by DLRMs. Consequently, we achieved INT4 quantization of DLRM models without any accuracy drop. We further propose two techniques that improve and accelerate the conventional QAT workload specifically for the embedding tables in the recommendation models. Furthermore, to achieve efficient training, we quantize the gradients of the embedding tables into INT8 on top of the well-supported specified sparsification. We show that combining gradient sparsification and quantization together significantly reduces the amount of communication. Briefly, DQRM models with INT4 can achieve 79.07% accuracy on Kaggle with 0.27 GB model size, and 81.21% accuracy on the Terabyte dataset with 1.57 GB, which even outperform FP32 DLRMs that have much larger model sizes (2.16 GB on Kaggle and 12.58 on Terabyte).

FP8 training has emerged as a promising method for improving training efficiency. Existing frameworks accelerate training by applying FP8 computation to linear layers while leaving optimizer states and activations in higher precision, which fails to fully optimize memory usage. This paper introduces COAT (Compressing Optimizer States and Activations for FP8 Training), a novel FP8 training framework designed to significantly reduce memory footprint when training large models. COAT addresses current limitations through two key innovations: (1) Dynamic Range Expansion, which aligns optimizer state distributions more closely with the FP8 representation range, thereby reducing quantization error, and (2) Mixed-Granularity Activation Quantization, which optimizes activation memory using a combination of per-tensor and per-group quantization strategies. Experiments demonstrate that COAT effectively reduces end-to-end training memory footprint by 1.54x compared to BF16 while achieving nearly lossless performance across various tasks, such as Large Language Model pretraining and fine-tuning and Vision Language Model training. COAT also achieves a 1.43x end-to-end training speedup compared to BF16, performing on par with or surpassing TransformerEngine's speedup. COAT enables efficient full-parameter training of large models on fewer GPUs, and facilitates doubling the batch size in distributed training settings, providing a practical solution for scaling large-scale model training. The code is available at https://github.com/NVlabs/COAT., Comment: 16 pages. 9 Figures. 8 Tables

There has been a long history of women innovators producing outstanding contributions to society and public benefit yet having their work passed over or sidelined or attributed to male colleagues. This phenomenon has been coined the Matilda Effect. The amendments to the record of human achievements are now taking place, with an increasing pace in recent times due to greater social enlightenment and awareness and the interest in social justice. However, there remains a gap that must be addressed. In this article, we demonstrate the disparity in scientific recognition for a handful of case studies through search data collected via Google Trends and plotted as time-series figures and choropleth maps. Search trends reflect a noticeable divergence between recognition of female and male innovators. However, we note that in more well-known cases of the Matilda Effect, such as the historical account of Rosalind Franklin vs. James Watson and Francis Crick, the differences become less pronounced, emphasizing the importance of publicizing recognition. In response to this revelation, this article presents the stories of several women innovators and their great achievements. We identify the truth behind several discoveries and inventions, while revealing the full nature of this historical problem of social exclusion.

Preclinical diffusion MRI (dMRI) has proven value in methods development and validation, characterizing the biological basis of diffusion phenomena, and comparative anatomy. While dMRI enables in vivo non-invasive characterization of tissue, ex vivo dMRI is increasingly being used to probe tissue microstructure and brain connectivity. Ex vivo dMRI has several experimental advantages that facilitate high spatial resolution and high signal-to-noise ratio (SNR) images, cutting-edge diffusion contrasts, and direct comparison with histological data as a methodological validation. However, there are a number of considerations that must be made when performing ex vivo experiments. The steps from tissue preparation, image acquisition and processing, and interpretation of results are complex, with many decisions that not only differ dramatically from in vivo imaging of small animals, but ultimately affect what questions can be answered using the data. This work concludes a 3-part series of recommendations and considerations for preclinical dMRI. Herein, we describe best practices for dMRI of ex vivo tissue, with a focus on image pre-processing, data processing and model fitting, and tractography. In each section, we attempt to provide guidelines and recommendations, but also highlight areas for which no guidelines exist (and why), and where future work should lie. We end by providing guidelines on code sharing and data sharing, and point towards open-source software and databases specific to small animal and ex vivo imaging., Comment: Part 3 of 3 in "Considerations and recommendations for preclinical diffusion MRI"

We explore the fair distribution of a set of $m$ indivisible chores among $n$ agents, where each agent's costs are evaluated using a monotone cost function. Our focus lies on two fairness criteria: envy-freeness up to any item (EFX) and a relaxed notion, namely envy-freeness up to the transfer of any item (tEFX). We demonstrate that a 2-approximate EFX allocation exists and is computable in polynomial time for three agents with subadditive cost functions, improving upon the previous $(2 + \sqrt{6})$ approximation for additive cost functions. This result requires extensive case analysis. Christoforidis et al. (IJCAI'24) independently claim the same approximation for additive cost functions; however, we provide a counter-example to their algorithm. We expand the number of agents to any number to get the same approximation guarantee with the assumption of partially identical ordering (IDO) for the cost functions. Additionally, we establish that a tEFX allocation is achievable for three agents if one has an additive 2-ratio bounded cost function, while the others may have general monotone cost functions. This is an improvement from the prior requirement of two agents with additive 2-ratio bounded cost functions. This allocation can also be extended to agent groups with identical valuations. Further, we show various analyses of EFX allocations for chores, such as the relaxations for additive $\alpha$-ratio-bounded cost functions.

In $D$ dimensional de Sitter space, a scalar field has an infinite tower of special tachyonic mass values at which enhanced shift symmetries appear. After modding out by these shift symmetries, these fields correspond to the unitary irreducible representations of the de Sitter group known as the discrete series. We show that in $D=2$ these theories have global conformal symmetry. In all but the massless case, these theories have no stress tensor and the conformal symmetry does not act in the usual way on the scalar field. We find the conformal symmetry by explicitly computing the correlators of the shift invariant local operators and showing that they take conformally invariant forms. We also demonstrate how these fields are self dual in $D=2$, and dual to the shift invariant massive vector fields, which are therefore also conformally invariant., Comment: 30 pages

We propose PixelGaussian, an efficient feed-forward framework for learning generalizable 3D Gaussian reconstruction from arbitrary views. Most existing methods rely on uniform pixel-wise Gaussian representations, which learn a fixed number of 3D Gaussians for each view and cannot generalize well to more input views. Differently, our PixelGaussian dynamically adapts both the Gaussian distribution and quantity based on geometric complexity, leading to more efficient representations and significant improvements in reconstruction quality. Specifically, we introduce a Cascade Gaussian Adapter to adjust Gaussian distribution according to local geometry complexity identified by a keypoint scorer. CGA leverages deformable attention in context-aware hypernetworks to guide Gaussian pruning and splitting, ensuring accurate representation in complex regions while reducing redundancy. Furthermore, we design a transformer-based Iterative Gaussian Refiner module that refines Gaussian representations through direct image-Gaussian interactions. Our PixelGaussian can effectively reduce Gaussian redundancy as input views increase. We conduct extensive experiments on the large-scale ACID and RealEstate10K datasets, where our method achieves state-of-the-art performance with good generalization to various numbers of views. Code: https://github.com/Barrybarry-Smith/PixelGaussian., Comment: Code is available at: https://github.com/Barrybarry-Smith/PixelGaussian

We consider the fundamental problem of fairly allocating a set of indivisible items among agents having valuations that are represented by a multi-graph -- here, agents appear as the vertices and items as the edges between them and each vertex (agent) only values the set of its incident edges (items). The goal is to find a fair, i.e., envy-free up to any item (EFX) allocation. This model has recently been introduced by Christodoulou et al. (EC'23) where they show that EFX allocations always exist on simple graphs for monotone valuations, i.e., where any two agents can share at most one edge (item). A natural question arises as to what happens when we go beyond simple graphs and study various classes of multi-graphs? We answer the above question affirmatively for the valuation class of bipartite multi-graphs and multi-cycles. Our main positive result is that EFX allocations on bipartite multi-graphs (and multi-cycles) always exist and can be computed in polynomial time for additive valuations. We, therefore, push the frontiers of our understanding of EFX allocations and expand the scenarios where they are known to exist for an arbitrary number of agents. Next, we study EFX orientations (i.e., allocations where every item is allocated to one of its two endpoint agents) and give a complete picture of when they exist for bipartite multi-graphs dependent on two parameters -- the number of edges shared between any two agents and the diameter of the graph. Finally, we prove that it is NP-complete to determine whether a given fair division instance on a bipartite multi-graph admits an EFX orientation.

We propose GE2E-KWS -- a generalized end-to-end training and evaluation framework for customized keyword spotting. Specifically, enrollment utterances are separated and grouped by keywords from the training batch and their embedding centroids are compared to all other test utterance embeddings to compute the loss. This simulates runtime enrollment and verification stages, and improves convergence stability and training speed by optimizing matrix operations compared to SOTA triplet loss approaches. To benchmark different models reliably, we propose an evaluation process that mimics the production environment and compute metrics that directly measure keyword matching accuracy. Trained with GE2E loss, our 419KB quantized conformer model beats a 7.5GB ASR encoder by 23.6% relative AUC, and beats a same size triplet loss model by 60.7% AUC. Our KWS models are natively streamable with low memory footprints, and designed to continuously run on-device with no retraining needed for new keywords (zero-shot)., Comment: 8 pages, 6 figures, 2 tables The paper is accepted in IEEE Spoken Language Technology (SLT) 2024

Synthetic data generation faces significant challenges in accurately replicating real data, particularly with tabular data, where achieving high fidelity and utility is critical. While numerous methods have been developed, the most effective approach for creating high-quality synthetic data for network traffic security remains to be seen. This study conducts a comprehensive comparative analysis of non-AI, conventional AI, and generative AI techniques for synthetic tabular data generation using two widely recognized cybersecurity datasets: NSL-KDD and CICIDS-2017. Particular emphasis was placed on prominent GAN models for tabular data generation, including CTGAN, CopulaGAN, GANBLR++, and CastGAN. The results indicate that GAN-based methods, particularly CTGAN and CopulaGAN, outperform non-AI and conventional AI approaches in terms of fidelity and utility. To the best of our knowledge, this research contributes to the field by offering the first comparative evaluation of these methods specifically for cybersecurity network traffic data, filling a critical gap in the literature. It also introduces mutual information for feature selection, further enhancing the quality of the generated synthetic data. These findings provide valuable guidance for researchers seeking the most suitable synthetic data generation method in cybersecurity applications.

We present, for the first time, power splitters with multiple channel configurations in one-dimensional grating waveguides (1DGWs) that maintain crystal lattice-sensitive Bloch mode profiles without perturbation across all output channels, all within an ultra-miniaturized footprint of just 2.1 x 2.2 {\mu}m^2. This novel capability reduces the need for transition regions, simplifies multi-channel configurations of 1DGWs, and maximizes the effective use of chip area. The pixelated metamaterial approach, integrated with a time-domain heuristic algorithm, is utilized to concurrently achieve broadband operation, optimized dispersion control, and minimal loss. We experimentally demonstrate that our 1x2 and 1x3 metasplitters achieve average minimum losses per channel of 3.80 dB and 5.36 dB, respectively, just 0.80 dB and 0.59 dB above ideal splitting. The measurements for both designs demonstrate a 1 dB bandwidth of 15 nm, with excellent uniformity across all output channels. These versatile metasplitter designs can serve as fundamental building blocks for ultrahigh-bandwidth, densely integrated photonic circuits and in scenarios where slow light is essential., Comment: 9 pages, 3 figures, 7 pages supplementary information

Vision-centric autonomous driving has demonstrated excellent performance with economical sensors. As the fundamental step, 3D perception aims to infer 3D information from 2D images based on 3D-2D projection. This makes driving perception models susceptible to sensor configuration (e.g., camera intrinsics and extrinsics) variations. However, generalizing across camera configurations is important for deploying autonomous driving models on different car models. In this paper, we present UniDrive, a novel framework for vision-centric autonomous driving to achieve universal perception across camera configurations. We deploy a set of unified virtual cameras and propose a ground-aware projection method to effectively transform the original images into these unified virtual views. We further propose a virtual configuration optimization method by minimizing the expected projection error between original cameras and virtual cameras. The proposed virtual camera projection can be applied to existing 3D perception methods as a plug-and-play module to mitigate the challenges posed by camera parameter variability, resulting in more adaptable and reliable driving perception models. To evaluate the effectiveness of our framework, we collect a dataset on Carla by driving the same routes while only modifying the camera configurations. Experimental results demonstrate that our method trained on one specific camera configuration can generalize to varying configurations with minor performance degradation., Comment: Preprint; 14 pages, 5 figures, 2 tables; Code at https://github.com/ywyeli/UniDrive

Ultrafast pulses, particularly those with durations under 100 femtoseconds, are crucial in achieving unprecedented precision and control in light-matter interactions. However, conventional on-chip photonic platforms are not inherently designed for ultrafast time-domain operations, posing a significant challenge in achieving essential parameters such as high peak power and high temporal resolution. This challenge is particularly pronounced when propagating through dispersive integrated waveguides, unlike classical applications where dispersion is typically near-zero or linear. In addressing this challenge, we present a design methodology for ultrafast pulse propagation in dispersive integrated waveguides, specifically focused on enhancing the time-domain characteristics of one-dimensional grating waveguides (1DGWs). The proposed methodology aims to determine the optimal structural parameters for achieving maximum peak power, enhanced temporal resolution, and extended pulse storage duration during ultrafast pulse propagation. To validate this approach, we design and fabricate two specialized 1DGWs on a silicon-on-insulator (SOI) platform. A digital finite impulse response (FIR) model, trained with both transmission and phase measurement data, is employed to obtain ultrafast time-domain characteristics, enabling easy extraction of these results. Our approach achieves a 2.78-fold increase in peak power and reduces pulse broadening by 24%, resulting in a smaller sacrifice in temporal resolution. These results can possibly pave the way for advanced light-matter interactions within dispersive integrated waveguides., Comment: 19 pages, 5 figures

Fluid transport across nanometric channels induced by electric, pressure and concentration gradients is ubiquitous in biological systems and fosters various applications. In this context, computer simulation setups with well-defined open-boundary equilibrium starting states are essential in understanding and assisting experimental studies. However, open-boundary computational methods are scarce and typically do not satisfy all the equilibrium conditions imposed by reality. Namely, in the absence of external gradients, 1) the system of interest (SoI) must be at thermodynamic and chemical equilibrium with an infinite reservoir of particles, 2) the fluctuations of the SoI in equilibrium should sample the grand canonical ensemble, 3) the local solvation thermodynamics, which is extremely sensitive to finite-size effects due to solvent depletion, should be correctly described. This point is particularly relevant for out-of-equilibrium systems. Finally, 4) the method should be robust enough to deal with phase transitions and coexistence conditions in the SoI. In this study, we demonstrate with prototypical liquid systems embedded into a reservoir of ideal gas particles that the adaptive resolution simulation (AdResS) method, coupled with particle insertion/deletion steps, satisfies all these requirements. Therefore, this AdResS setup is suitable for performing equilibrium and non-equilibrium simulations of open systems.

Optimal Bayesian decision making on what geoscientific data to acquire requires stating a prior model of uncertainty. Data acquisition is then optimized by reducing uncertainty on some property of interest maximally, and on average. In the context of exploration, very few, sometimes no data at all, is available prior to data acquisition planning. The prior model therefore needs to include human interpretations on the nature of spatial variability, or on analogue data deemed relevant for the area being explored. In mineral exploration, for example, humans may rely on conceptual models on the genesis of the mineralization to define multiple hypotheses, each representing a specific spatial variability of mineralization. More often than not, after the data is acquired, all of the stated hypotheses may be proven incorrect, i.e. falsified, hence prior hypotheses need to be revised, or additional hypotheses generated. Planning data acquisition under wrong geological priors is likely to be inefficient since the estimated uncertainty on the target property is incorrect, hence uncertainty may not be reduced at all. In this paper, we develop an intelligent agent based on partially observable Markov decision processes that plans optimally in the case of multiple geological or geoscientific hypotheses on the nature of spatial variability. Additionally, the artificial intelligence is equipped with a method that allows detecting, early on, whether the human stated hypotheses are incorrect, thereby saving considerable expense in data acquisition. Our approach is tested on a sediment-hosted copper deposit, and the algorithm presented has aided in the characterization of an ultra high-grade deposit in Zambia in 2023.

Several quantum radar concepts have been proposed that exploit the entanglement found in two-mode squeezed vacuum states of the electromagnetic field, the most prominent being radar based on quantum illumination. Classical radars are sometimes required to distinguish between true echos of their transmitted signals and signals generated by interferors or spoofers. How vulnerable to spoofing is quantum illumination? We analyze the scenario of a radar operator trying to detect the presence of a classical spoofer employing a measure-and-prepare strategy against a quantum radar. We consider two spoofing strategies - (1) direct detection and number state preparation, and (2) heterodyne detection and coherent state preparation. In each case, the radar operator performs a hypothesis test to decide if received pulses are true returns or spoofs. Since the spoofer can not reproduce the entanglement with modes retained by the radar operator, both approaches to spoofing are to some extent detectable. We quantify the effectiveness of the spoof in terms of the fidelity between the real return and the spoof return, and the probability of error in spoof detection. We find that in the absence of noise and loss, direct detection tends to produce spoofs with greater fidelity, which are therefore harder to detect. Moreover, this advantage survives the introduction of noise and loss into the model. Our results suggest that entanglement is a novel resource available to quantum radar for detecting spoofing., Comment: 23 pages, 6 figures

We evaluate the impact of pretraining Graph Transformer architectures on atom-level quantum-mechanical features for the modeling of absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of drug-like compounds. We compare this pretraining strategy with two others: one based on molecular quantum properties (specifically the HOMO-LUMO gap) and one using a self-supervised atom masking technique. After fine-tuning on Therapeutic Data Commons ADMET datasets, we evaluate the performance improvement in the different models observing that models pretrained with atomic quantum mechanical properties produce in general better results. We then analyse the latent representations and observe that the supervised strategies preserve the pretraining information after finetuning and that different pretrainings produce different trends in latent expressivity across layers. Furthermore, we find that models pretrained on atomic quantum mechanical properties capture more low-frequency laplacian eigenmodes of the input graph via the attention weights and produce better representations of atomic environments within the molecule. Application of the analysis to a much larger non-public dataset for microsomal clearance illustrates generalizability of the studied indicators. In this case the performances of the models are in accordance with the representation analysis and highlight, especially for the case of masking pretraining and atom-level quantum property pretraining, how model types with similar performance on public benchmarks can have different performances on large scale pharmaceutical data.

In vision-language models (VLMs), visual tokens usually consume a significant amount of computational overhead, despite their sparser information density compared to text tokens. To address this, most existing methods learn a network to prune redundant visual tokens and require additional training data. Differently, we propose an efficient training-free token optimization mechanism dubbed SparseVLM without extra parameters or fine-tuning costs. Concretely, given that visual tokens complement text tokens in VLMs for linguistic reasoning, we select visual-relevant text tokens to rate the significance of vision tokens within the self-attention matrix extracted from the VLMs. Then we progressively prune irrelevant tokens. To maximize sparsity while retaining essential information, we introduce a rank-based strategy to adaptively determine the sparsification ratio for each layer, alongside a token recycling method that compresses pruned tokens into more compact representations. Experimental results show that our SparseVLM improves the efficiency of various VLMs across a range of image and video understanding tasks. In particular, LLaVA equipped with SparseVLM reduces 61% to 67% FLOPs with a compression ratio of 78% while maintaining 93% of the accuracy. Our code is available at https://github.com/Gumpest/SparseVLMs., Comment: 17 pages

Anomaly detection is an important problem with applications in various domains such as fraud detection, pattern recognition or medical diagnosis. Several algorithms have been introduced using classical computing approaches. However, using quantum computing for solving anomaly detection problems in time series data is a widely unexplored research field. This paper explores the application of quantum autoencoders to time series anomaly detection. We investigate two primary techniques for classifying anomalies: (1) Analyzing the reconstruction error generated by the quantum autoencoder and (2) latent representation analysis. Our simulated experimental results, conducted across various ansaetze, demonstrate that quantum autoencoders consistently outperform classical deep learning-based autoencoders across multiple datasets. Specifically, quantum autoencoders achieve superior anomaly detection performance while utilizing 60-230 times fewer parameters and requiring five times fewer training iterations. In addition, we implement our quantum encoder on real quantum hardware. Our experimental results demonstrate that quantum autoencoders achieve anomaly detection performance on par with their simulated counterparts., Comment: 22 pages, 16 figures