PyTorchKR Papers

최신 논문

총 139,333편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2505.13567공식 코드 · Papers with Code

Learning Dynamics of RNNs in Closed-Loop Environments

Recurrent neural networks (RNNs) trained on neuroscience-inspired tasks offer powerful models of brain computation. However, typical training paradigms rely on open-loop, supervised settings, whereas real-world learning unfolds in closed-loop environments. Here, we develop a mathematical theory describing the learning dynamics of linear RNNs trained in closed-loop contexts. We first demonstrate that two otherwise identical RNNs, trained in either closed- or open-loop modes, follow markedly different learning trajectories. To probe this divergence, we analytically characterize the closed-loop case, revealing distinct stages aligned with the evolution of the training loss. Specifically, we show that the learning dynamics of closed-loop RNNs, in contrast to open-loop ones, are governed by an interplay between two competing objectives: short-term policy improvement and long-term stability of the agent-environment interaction. Finally, we apply our framework to a realistic motor control task, highlighting its broader applicability. Taken together, our results underscore the importance of modeling closed-loop dynamics in a biologically plausible setting.

arXiv:2505.13534공식 코드 · Papers with Code

InterFeat: An Automated Pipeline for Finding Interesting Hypotheses in Structured Biomedical Data

Finding interesting phenomena is the core of scientific discovery, but it is a manual, ill-defined concept. We present an integrative pipeline for automating the discovery of interesting simple hypotheses (feature-target relations with effect direction and a potential underlying mechanism) in structured biomedical data. The pipeline combines machine learning, knowledge graphs, literature search and Large Language Models. We formalize "interestingness" as a combination of novelty, utility and plausibility. On 8 major diseases from the UK Biobank, our pipeline consistently recovers risk factors years before their appearance in the literature. 40--53% of our top candidates were validated as interesting, compared to 0--7% for a SHAP-based baseline. Overall, 28% of 109 candidates were interesting to medical experts. The pipeline addresses the challenge of operationalizing "interestingness" scalably and for any target. We release data and code: https://github.com/LinialLab/InterFeat

arXiv:2505.12653공식 코드 · Papers with Code

High-dimensional structure underlying individual differences in naturalistic visual experience

How do different brains create unique visual experiences from identical sensory input? While neural representations vary across individuals, the fundamental architecture underlying these differences remains poorly understood. Here, we reveal that individual visual experience emerges from a high-dimensional neural geometry across the visual cortical hierarchy. Using spectral decomposition of fMRI responses during naturalistic movie viewing, we find that idiosyncratic neural patterns persist across multiple orders of magnitude of latent dimensions. Remarkably, each dimensional range encodes qualitatively distinct aspects of individual processing, and this multidimensional neural geometry predicts subsequent behavioral differences in memory recall. These fine-grained patterns of inter-individual variability cannot be reduced to those detected by conventional intersubject correlation measures. Our findings demonstrate that subjective visual experience arises from information integrated across an expansive multidimensional manifold. This geometric framework offers a powerful new lens for understanding how diverse brains construct unique perceptual worlds from shared experiences.

arXiv:2505.11538공식 코드 · Papers with Code

BrainNetMLP: An Efficient and Effective Baseline for Functional Brain Network Classification

Recent studies have made great progress in functional brain network classification by modeling the brain as a network of Regions of Interest (ROIs) and leveraging their connections to understand brain functionality and diagnose mental disorders. Various deep learning architectures, including Convolutional Neural Networks, Graph Neural Networks, and the recent Transformer, have been developed. However, despite the increasing complexity of these models, the performance gain has not been as salient. This raises a question: Does increasing model complexity necessarily lead to higher classification accuracy? In this paper, we revisit the simplest deep learning architecture, the Multi-Layer Perceptron (MLP), and propose a pure MLP-based method, named BrainNetMLP, for functional brain network classification, which capitalizes on the advantages of MLP, including efficient computation and fewer parameters. Moreover, BrainNetMLP incorporates a dual-branch structure to jointly capture both spatial connectivity and spectral information, enabling precise spatiotemporal feature fusion. We evaluate our proposed BrainNetMLP on two public and popular brain network classification datasets, the Human Connectome Project (HCP) and the Autism Brain Imaging Data Exchange (ABIDE). Experimental results demonstrate pure MLP-based methods can achieve state-of-the-art performance, revealing the potential of MLP-based models as more efficient yet effective alternatives in functional brain network classification. The code will be available at https://github.com/JayceonHo/BrainNetMLP.

arXiv:2505.11384공식 코드 · Papers with Code

MOSAIK: Multi-Origin Spatial Transcriptomics Analysis and Integration Kit

Spatial transcriptomics (ST) has revolutionised transcriptomics analysis by preserving tissue architecture, allowing researchers to study gene expression in its native spatial context. However, despite its potential, ST still faces significant technical challenges. Two major issues include: (1) the integration of raw data into coherent and reproducible analysis workflows, and (2) the accurate assignment of transcripts to individual cells. To address these challenges, we present MOSAIK, the first fully integrated, end-to-end workflow that supports raw data from both NanoString CosMx Spatial Molecular Imager (CosMx) and 10x Genomics Xenium In Situ (Xenium). MOSAIK (Multi-Origin Spatial Transcriptomics Analysis and Integration Kit) unifies transcriptomics and imaging data into a single Python object based on the spatialdata format. This unified structure ensures compatibility with a broad range of Python tools, enabling robust quality control and downstream analyses. With MOSAIK, users can perform advanced analyses such as re-segmentation (to more accurately assign transcripts to individual cells), cell typing, tissue domain identification, and cell-cell communication within a seamless and reproducible Python environment.

arXiv:2505.14402공식 코드 · Papers with Code

OmniGenBench: A Modular Platform for Reproducible Genomic Foundation Models Benchmarking

The code of nature, embedded in DNA and RNA genomes since the origin of life, holds immense potential to impact both humans and ecosystems through genome modeling. Genomic Foundation Models (GFMs) have emerged as a transformative approach to decoding the genome. As GFMs scale up and reshape the landscape of AI-driven genomics, the field faces an urgent need for rigorous and reproducible evaluation. We present OmniGenBench, a modular benchmarking platform designed to unify the data, model, benchmarking, and interpretability layers across GFMs. OmniGenBench enables standardized, one-command evaluation of any GFM across five benchmark suites, with seamless integration of over 31 open-source models. Through automated pipelines and community-extensible features, the platform addresses critical reproducibility challenges, including data transparency, model interoperability, benchmark fragmentation, and black-box interpretability. OmniGenBench aims to serve as foundational infrastructure for reproducible genomic AI research, accelerating trustworthy discovery and collaborative innovation in the era of genome-scale modeling.

arXiv:2505.14017공식 코드 · Papers with Code

End-to-end Cortical Surface Reconstruction from Clinical Magnetic Resonance Images

Surface-based cortical analysis is valuable for a variety of neuroimaging tasks, such as spatial normalization, parcellation, and gray matter (GM) thickness estimation. However, most tools for estimating cortical surfaces work exclusively on scans with at least 1 mm isotropic resolution and are tuned to a specific magnetic resonance (MR) contrast, often T1-weighted (T1w). This precludes application using most clinical MR scans, which are very heterogeneous in terms of contrast and resolution. Here, we use synthetic domain-randomized data to train the first neural network for explicit estimation of cortical surfaces from scans of any contrast and resolution, without retraining. Our method deforms a template mesh to the white matter (WM) surface, which guarantees topological correctness. This mesh is further deformed to estimate the GM surface. We compare our method to recon-all-clinical (RAC), an implicit surface reconstruction method which is currently the only other tool capable of processing heterogeneous clinical MR scans, on ADNI and a large clinical dataset (n=1,332). We show a approximately 50 % reduction in cortical thickness error (from 0.50 to 0.24 mm) with respect to RAC and better recovery of the aging-related cortical thinning patterns detected by FreeSurfer on high-resolution T1w scans. Our method enables fast and accurate surface reconstruction of clinical scans, allowing studies (1) with sample sizes far beyond what is feasible in a research setting, and (2) of clinical populations that are difficult to enroll in research studies. The code is publicly available at https://github.com/simnibs/brainnet.

arXiv:2505.13940공식 코드 · Papers with Code

DrugPilot: LLM-based Parameterized Reasoning Agent for Drug Discovery

In the field of AI4Science, large-scale language models (LLMs) show great potential to parse complex scientific semantics, integrate cross-disciplinary knowledge, and assist critical task research. However, in the field of drug discovery, despite the optimization through professional data pre-training, context window expansion, and internet search, the existing LLMs are still facing challenges such as massive multi-modal and heterogeneous data processing, domain knowledge dynamic updating delay, and insufficient confidence in predicting the results of complex computational tasks. To address these challenges, we propose the DrugPilot, an LLM-based agent with parameterized reasoning for drug discovery. DrugPilot addresses key limitations of traditional end-to-end LLM prediction approaches through its parametric inference architecture. This agent system supports major phases of the drug discovery pipeline, facilitating automated planning and execution of multi-stage research tasks. To address the critical challenge of multi-modal drug data analysis (incorporating both public datasets and user-submitted data), we developed an interactive parameterized memory pool. This innovative component standardizes real-world drug data into parametric representations, simultaneously enabling efficient knowledge retrieval in multi-turn dialogue while mitigating the information loss inherent in text-based data transmission. Additionally, we created a drug instruct dataset across 8 essential drug discovery tasks for model fine-tuning and evaluation. Based on the Berkeley function calling evaluation framework, DrugPilot demonstrated the most advanced tool calling capabilities on our drug discovery tool instruction dataset, outperforming existing agents (e.g., ReAct, LoT). Specifically, it achieves task completion rates of 98.0%, 93.5%, and 64.0% on simple, multiple, and multi-turn tasks, respectively.

arXiv:2505.13723공식 코드 · Papers with Code

Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project

Gaussian processes (GPs) play an essential role in biostatistics, scientific machine learning, and Bayesian optimization for their ability to provide probabilistic predictions and model uncertainty. However, GP inference struggles to scale to large datasets (which are common in modern applications), since it requires the solution of a linear system whose size scales quadratically with the number of samples in the dataset. We propose an approximate, distributed, accelerated sketch-and-project algorithm ($\texttt{ADASAP}$) for solving these linear systems, which improves scalability. We use the theory of determinantal point processes to show that the posterior mean induced by sketch-and-project rapidly converges to the true posterior mean. In particular, this yields the first efficient, condition number-free algorithm for estimating the posterior mean along the top spectral basis functions, showing that our approach is principled for GP inference. $\texttt{ADASAP}$ outperforms state-of-the-art solvers based on conjugate gradient and coordinate descent across several benchmark datasets and a large-scale Bayesian optimization task. Moreover, $\texttt{ADASAP}$ scales to a dataset with $> 3 \cdot 10^8$ samples, a feat which has not been accomplished in the literature.

arXiv:2505.13373공식 코드 · Papers with Code

State- versus Reaction-Based Information Processing in Biochemical Networks

Trajectory mutual information is frequently used to quantify information transfer in biochemical systems. Tractable solutions of the trajectory mutual information can be obtained via the widely used Linear-Noise Approximation (LNA) using Gaussian channel theory. This approach is expected to be accurate for sufficiently large systems. However, recent observations show that there are cases, where the mutual information obtained this way differs qualitatively from results derived using an exact Markov jump process formalism, and that the differences persist even for large systems. In this letter, we show that these differences can be explained by introducing the notion of reaction- versus state-based descriptions of trajectories. In chemical systems, the information is encoded in the sequence of reaction events, and the reaction-based trajectories of Markov jump processes capture this information. In contrast, the commonly used form of the LNA uses a state (concentration) based description of trajectories, which contains, in general, less information than a reaction-based description. Here, we show that an alternative formulation of the LNA that retains the reaction-specific information of trajectories can accurately describe the trajectory mutual information for large systems. We illustrate the consequences of different trajectory descriptions for two common cellular reaction motifs and discuss the connection with Berg-Purcell and Maximum-Likelihood sensing.

arXiv:2505.11853공식 코드 · Papers with Code

Measurement Score-Based Diffusion Model

Diffusion models are widely used in applications ranging from image generation to inverse problems. However, training diffusion models typically requires clean ground-truth images, which are unavailable in many applications. We introduce the Measurement Score-based diffusion Model (MSM), a novel framework that learns partial measurement scores using only noisy and subsampled measurements. MSM models the distribution of full measurements as an expectation over partial scores induced by randomized subsampling. To make the MSM representation computationally efficient, we also develop a stochastic sampling algorithm that generates full images by using a randomly selected subset of partial scores at each step. We additionally propose a new posterior sampling method for solving inverse problems that reconstructs images using these partial scores. We provide a theoretical analysis that bounds the Kullback-Leibler divergence between the distributions induced by full and stochastic sampling, establishing the accuracy of the proposed algorithm. We demonstrate the effectiveness of MSM on natural images and multi-coil MRI, showing that it can generate high-quality images and solve inverse problems -- all without access to clean training data. Code is available at https://github.com/wustl-cig/MSM.

arXiv:2505.11618공식 코드 · Papers with Code

Benchmarking Spatiotemporal Reasoning in LLMs and Reasoning Models: Capabilities and Challenges

Spatiotemporal reasoning plays a key role in Cyber-Physical Systems (CPS). Despite advances in Large Language Models (LLMs) and Large Reasoning Models (LRMs), their capacity to reason about complex spatiotemporal signals remains underexplored. This paper proposes a hierarchical SpatioTemporal reAsoning benchmaRK, STARK, to systematically evaluate LLMs across three levels of reasoning complexity: state estimation (e.g., predicting field variables, localizing and tracking events in space and time), spatiotemporal reasoning over states (e.g., inferring spatial-temporal relationships), and world-knowledge-aware reasoning that integrates contextual and domain knowledge (e.g., intent prediction, landmark-aware navigation). We curate 26 distinct spatiotemporal tasks with diverse sensor modalities, comprising 14,552 challenges where models answer directly or by Python Code Interpreter. Evaluating 3 LRMs and 8 LLMs, we find LLMs achieve limited success in tasks requiring geometric reasoning (e.g., multilateration or triangulation), particularly as complexity increases. Surprisingly, LRMs show robust performance across tasks with various levels of difficulty, often competing or surpassing traditional first-principle-based methods. Our results show that in reasoning tasks requiring world knowledge, the performance gap between LLMs and LRMs narrows, with some LLMs even surpassing LRMs. However, the LRM o3 model continues to achieve leading performance across all evaluated tasks, a result attributed primarily to the larger size of the reasoning models. STARK motivates future innovations in model architectures and reasoning paradigms for intelligent CPS by providing a structured framework to identify limitations in the spatiotemporal reasoning of LLMs and LRMs.

arXiv:2505.11580공식 코드 · Papers with Code

Flash Invariant Point Attention

Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-training generative models without length restrictions and generating structures of thousands of residues. FlashIPA is available at https://github.com/flagshippioneering/flash_ipa.

arXiv:2505.11241공식 코드 · Papers with Code

Unique global solution of an integral-differential equation of Footloose Entrepreneur model in new economic geography

This paper studies the Footloose Entrepreneur model in new economic geography in continuous space. In an appropriate function space, the model is formulated as an initial value problem for an infinite-dimensional ordinary differential equation. A unique global solution is constructed based on the Banach fixed point theorem. The stability of a homogeneous stationary solution is then investigated and numerical simulations of the asymptotic behavior of the solution are performed. Numerical solutions starting near the unstable homogeneous stationary solution converge to spike-shaped stationary solutions, and the number of spikes decreases with decreasing transport costs and strengthening preference for variety.

arXiv:2505.10793공식 코드 · Papers with Code

SongEval: A Benchmark Dataset for Song Aesthetics Evaluation

Aesthetics serve as an implicit and important criterion in song generation tasks that reflect human perception beyond objective metrics. However, evaluating the aesthetics of generated songs remains a fundamental challenge, as the appreciation of music is highly subjective. Existing evaluation metrics, such as embedding-based distances, are limited in reflecting the subjective and perceptual aspects that define musical appeal. To address this issue, we introduce SongEval, the first open-source, large-scale benchmark dataset for evaluating the aesthetics of full-length songs. SongEval includes over 2,399 songs in full length, summing up to more than 140 hours, with aesthetic ratings from 16 professional annotators with musical backgrounds. Each song is evaluated across five key dimensions: overall coherence, memorability, naturalness of vocal breathing and phrasing, clarity of song structure, and overall musicality. The dataset covers both English and Chinese songs, spanning nine mainstream genres. Moreover, to assess the effectiveness of song aesthetic evaluation, we conduct experiments using SongEval to predict aesthetic scores and demonstrate better performance than existing objective evaluation metrics in predicting human-perceived musical quality.

arXiv:2505.14670공식 코드 · Papers with Code

Quantum Optimization via Gradient-Based Hamiltonian Descent

With rapid advancements in machine learning, first-order algorithms have emerged as the backbone of modern optimization techniques, owing to their computational efficiency and low memory requirements. Recently, the connection between accelerated gradient methods and damped heavy-ball motion, particularly within the framework of Hamiltonian dynamics, has inspired the development of innovative quantum algorithms for continuous optimization. One such algorithm, Quantum Hamiltonian Descent (QHD), leverages quantum tunneling to escape saddle points and local minima, facilitating the discovery of global solutions in complex optimization landscapes. However, QHD faces several challenges, including slower convergence rates compared to classical gradient methods and limited robustness in highly non-convex problems due to the non-local nature of quantum states. Furthermore, the original QHD formulation primarily relies on function value information, which limits its effectiveness. Inspired by insights from high-resolution differential equations that have elucidated the acceleration mechanisms in classical methods, we propose an enhancement to QHD by incorporating gradient information, leading to what we call gradient-based QHD. Gradient-based QHD achieves faster convergence and significantly increases the likelihood of identifying global solutions. Numerical simulations on challenging problem instances demonstrate that gradient-based QHD outperforms existing quantum and classical methods by at least an order of magnitude.

arXiv:2505.14606공식 코드 · Papers with Code

Electrostatics from Laplacian Eigenbasis for Neural Network Interatomic Potentials

Recent advances in neural network interatomic potentials have emerged as a promising research direction. However, popular deep learning models often lack auxiliary constraints grounded in physical laws, which could accelerate training and improve fidelity through physics-based regularization. In this work, we introduce $\Phi$-Module, a universal plugin module that enforces Poisson's equation within the message-passing framework to learn electrostatic interactions in a self-supervised manner. Specifically, each atom-wise representation is encouraged to satisfy a discretized Poisson's equation, making it possible to acquire a potential $\boldsymbol{\phi}$ and a corresponding charge density $\boldsymbol{\rho}$ linked to the learnable Laplacian eigenbasis coefficients of a given molecular graph. We then derive an electrostatic energy term, crucial for improved total energy predictions. This approach integrates seamlessly into any existing neural potential with insignificant computational overhead. Experiments on the OE62 and MD22 benchmarks confirm that models combined with $\Phi$-Module achieve robust improvements over baseline counterparts. For OE62 error reduction ranges from 4.5\% to 17.8\%, and for MD22, baseline equipped with $\Phi$-Module achieves best results on 5 out of 14 cases. Our results underscore how embedding a first-principles constraint in neural interatomic potentials can significantly improve performance while remaining hyperparameter-friendly, memory-efficient and lightweight in training. Code will be available at \href{https://github.com/dunnolab/phi-module}{dunnolab/phi-module}.

arXiv:2505.14440공식 코드 · Papers with Code

Efficient Configuration-Constrained Tube MPC via Variables Restriction and Template Selection

Configuration-Constrained Tube Model Predictive Control (CCTMPC) offers flexibility by using a polytopic parameterization of invariant sets and the optimization of an associated vertex control law. This flexibility, however, often demands computational trade-offs between set parameterization accuracy and optimization complexity. This paper proposes two innovations that help the user tackle this trade-off. First, a structured framework is proposed, which strategically limits optimization degrees of freedom, significantly reducing online computation time while retaining stability guarantees. This framework aligns with Homothetic Tube MPC (HTMPC) under maximal constraints. Second, a template refinement algorithm that iteratively solves quadratic programs is introduced to balance polytope complexity and conservatism. Simulation studies on an illustrative benchmark problem as well as a high-dimensional ten-state system demonstrate the approach's efficiency, achieving robust performance with minimal computational overhead. The results validate a practical pathway to leveraging CCTMPC's adaptability without sacrificing real-time viability.

arXiv:2505.14245공식 코드 · Papers with Code

Path-integral molecular dynamics with actively-trained and universal machine learning force fields

Accounting for nuclear quantum effects (NQEs) can significantly alter material properties at finite temperatures. Atomic modeling using the path-integral molecular dynamics (PIMD) method can fully account for such effects, but requires computationally efficient and accurate models of interatomic interactions. Empirical potentials are fast but may lack sufficient accuracy, whereas quantum-mechanical calculations are highly accurate but computationally expensive. Machine-learned interatomic potentials offer a solution to this challenge, providing near-quantum-mechanical accuracy while maintaining high computational efficiency compared to density functional theory (DFT) calculations. In this context, an interface was developed to integrate moment tensor potentials (MTPs) from the MLIP-2 software package into PIMD calculations using the i-PI software package. This interface was then applied to active learning of potentials and to investigate the influence of NQEs on material properties, namely the temperature dependence of lattice parameters and thermal expansion coefficients, as well as radial distribution functions, for lithium hydride (LiH) and silicon (Si) systems. The results were compared with experimental data, quasi-harmonic approximation calculations, and predictions from the universal machine learning force field MatterSim. These comparisons demonstrated the high accuracy and effectiveness of the MTP-PIMD approach.

arXiv:2505.14001공식 코드 · Papers with Code

VeRecycle: Reclaiming Guarantees from Probabilistic Certificates for Stochastic Dynamical Systems after Change

Autonomous systems operating in the real world encounter a range of uncertainties. Probabilistic neural Lyapunov certification is a powerful approach to proving safety of nonlinear stochastic dynamical systems. When faced with changes beyond the modeled uncertainties, e.g., unidentified obstacles, probabilistic certificates must be transferred to the new system dynamics. However, even when the changes are localized in a known part of the state space, state-of-the-art requires complete re-certification, which is particularly costly for neural certificates. We introduce VeRecycle, the first framework to formally reclaim guarantees for discrete-time stochastic dynamical systems. VeRecycle efficiently reuses probabilistic certificates when the system dynamics deviate only in a given subset of states. We present a general theoretical justification and algorithmic implementation. Our experimental evaluation shows scenarios where VeRecycle both saves significant computational effort and achieves competitive probabilistic guarantees in compositional neural control.