AI 基础设施
AI 芯片、训练框架、推理优化、云服务等底层基础设施发展动态。
把20亿参数装进胸针？高通补齐了个人AI生态的最后一块拼图
高通技术公司正式发布了面向顶级智能手表的新一代旗舰平台——骁龙可穿戴平台至尊版。该平台集成了更强大的CPU、GPU和增强的终端侧AI引擎，旨在为下一代高端智能手表提供前所未有的性能、能效与连接能力，目标直指由苹果和三星主导的高端及专业运动健...
把20亿参数装进胸针？高通补齐了个人AI生态的最后一块拼图
高通技术公司正式发布了面向高端智能手表的全新旗舰平台——骁龙可穿戴平台至尊版。该平台在性能、能效与终端侧AI处理能力上进行了全面进化，旨在支持更先进的健康监测、独立通信功能，并为Wear OS生态提供强大硬件基础，预计将从2024年底开始引...
苹果春季新品奔着龙虾来了！AI性能暴涨8倍，8499元起
苹果在春季新品发布会上推出了多款产品，其AI性能相比前代提升了8倍。新品起售价为8499元人民币，标志着苹果在人工智能硬件领域的重大升级。
Who needs data centers in space when they can float offshore?
Offshore wind developer Aikido is deploying a subsea data center beneath a floating wind turbine, creating a carbon-neut...
沐曦股份：预计2026年第一季度净亏损9075.72万元-18151.43万元
中国高性能GPU设计公司沐曦股份发布2026年第一季度业绩预告，预计实现营业收入4亿元至6亿元，同比增长24.84%至87.26%。同时，公司预计净亏损为9075.72万元至1.82亿元，同比收窄21.93%至60.97%，显示出商业化加速...
特斯拉巨鲸廖凯原买入100万股英伟达，称还将继续买入
知名科技投资者、特斯拉早期重要股东廖凯原于3月4日透露，他在3日购入了100万股英伟达股票，并计划继续增持。他对此举的解释是确信“AI不是泡沫，现在只是开始”，这标志着其投资重心从特斯拉向AI算力领导者英伟达的显著调整。廖凯原将特斯拉重新定...
A Boundary Integral-based Neural Operator for Mesh Deformation
Researchers have developed a novel Boundary-Integral-based Neural Operator (BINO) for mesh deformation that formulates d...
A Boundary Integral-based Neural Operator for Mesh Deformation
Researchers developed the Boundary-Integral-based Neural Operator (BINO), a novel framework for mesh deformation that us...
A Boundary Integral-based Neural Operator for Mesh Deformation
Researchers have developed a Boundary-Integral-based Neural Operator (BINO) that uses a Dirichlet-type Green's tensor to...
A Boundary Integral-based Neural Operator for Mesh Deformation
Researchers have developed a Boundary-Integral-based Neural Operator (BINO) that combines boundary integral equations wi...
A Boundary Integral-based Neural Operator for Mesh Deformation
Researchers have developed the Boundary-Integral-based Neural Operator (BINO), a novel framework for mesh deformation th...
A Researcher's Guide to Empirical Risk Minimization
A new technical framework establishes a unified, modular approach for deriving high-probability regret bounds in Empiric...
A Researcher's Guide to Empirical Risk Minimization
A new technical guide presents a modular framework for deriving high-probability regret bounds in Empirical Risk Minimiz...
A Researcher's Guide to Empirical Risk Minimization
A new technical framework establishes high-probability regret bounds for Empirical Risk Minimization (ERM) through a mod...
A Researcher's Guide to Empirical Risk Minimization
Researchers have developed a comprehensive three-step framework for deriving high-probability regret bounds in Empirical...
A Researcher's Guide to Empirical Risk Minimization
A new technical framework establishes a unified, modular approach for deriving high-probability regret bounds in Empiric...
DRESS: A Continuous Framework for Structural Graph Refinement
The DRAGNN framework represents a breakthrough in graph isomorphism testing, empirically surpassing the distinguishing p...
DRESS: A Continuous Framework for Structural Graph Refinement
The DRESS framework is a novel family of algorithms that provides a scalable alternative to the Weisfeiler-Lehman hierar...
DRESS: A Continuous Framework for Structural Graph Refinement
The DRESS (DRess Edge State System) framework is a new family of algorithms that provides a computationally efficient al...
DRESS: A Continuous Framework for Structural Graph Refinement
The DRESS framework is a novel family of continuous dynamical system algorithms that provides a scalable alternative to ...
DRESS: A Continuous Framework for Structural Graph Refinement
The DRESS framework represents a novel approach to graph isomorphism testing using continuous dynamical systems instead ...
Stochastic Control Methods for Optimization
A novel stochastic control framework has been developed to solve global optimization problems for non-convex and non-dif...
Stochastic Control Methods for Optimization
A novel stochastic control framework provides a unified approach for finding global minima of complex, non-convex functi...
Stochastic Control Methods for Optimization
A novel stochastic control framework enables global optimization of non-convex, non-differentiable functions across Eucl...
Stochastic Control Methods for Optimization
A novel stochastic control framework enables global optimization across both Euclidean spaces and Wasserstein probabilit...
Stochastic Control Methods for Optimization
A new stochastic control framework reformulates static global optimization problems as dynamic stochastic control proble...
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
FAST is a novel framework for coreset selection that uses spectral graph theory and frequency-domain distribution matchi...
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
FAST is a novel AI framework that revolutionizes coreset selection for deep neural network training by formulating it as...
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
FAST is a novel DNN-free distribution-matching coreset selection framework that formulates the task as a graph-constrain...
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
FAST is a novel AI framework that revolutionizes deep neural network training efficiency through topology-aware frequenc...
FAST: Topology-Aware Frequency-Domain Distribution Matching for Coreset Selection
FAST is a novel AI framework that dramatically reduces computational and energy costs in deep learning model training by...
Physically Valid Biomolecular Interaction Modeling with Gauss-Seidel Projection
A new AI framework integrates a differentiable Gauss-Seidel projection module to enforce physical steric constraints in ...
Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances
Researchers have developed a novel regression-based method that drastically accelerates the computation of Wasserstein d...
Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances
A novel machine learning method enables fast estimation of Wasserstein distances by using a linear regression model trai...
Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances
Researchers have developed a fast linear regression method for estimating Wasserstein distances between probability dist...
Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances
Researchers have developed a novel regression-based method for fast, accurate estimation of Wasserstein distances betwee...
Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances
Researchers have developed a novel regression-based method for efficiently estimating Wasserstein distances between prob...
Nonparametric Reaction Coordinate Optimization with Histories: A Framework for Rare Event Dynamics
A novel nonparametric framework optimizes reaction coordinates for rare event analysis by incorporating full trajectory ...
Nonparametric Reaction Coordinate Optimization with Histories: A Framework for Rare Event Dynamics
A new nonparametric framework for optimizing reaction coordinates using trajectory histories enables robust analysis of ...
Covering Numbers for Deep ReLU Networks with Applications to Function Approximation and Nonparametric Regression
This research establishes the first comprehensive set of tight lower and upper bounds on the metric entropy of deep ReLU...
Covering Numbers for Deep ReLU Networks with Applications to Function Approximation and Nonparametric Regression
This study establishes the first tight lower and upper bounds on metric entropy for deep ReLU neural networks, quantifyi...
Covering Numbers for Deep ReLU Networks with Applications to Function Approximation and Nonparametric Regression
A landmark study establishes the first comprehensive, tight bounds on the metric entropy of deep ReLU networks, providin...
Covering Numbers for Deep ReLU Networks with Applications to Function Approximation and Nonparametric Regression
A new research paper provides the first comprehensive set of tight lower and upper bounds on the metric entropy (logarit...
A Normal Map-Based Proximal Stochastic Gradient Method: Convergence and Identification Properties
The Normal Map-based Proximal Stochastic Gradient Method (NSGD) is a novel algorithm that solves the long-standing limit...
A Normal Map-Based Proximal Stochastic Gradient Method: Convergence and Identification Properties
The Normal Map-based Proximal Stochastic Gradient Method (NSGD) is a novel optimization algorithm that addresses the key...
A Normal Map-Based Proximal Stochastic Gradient Method: Convergence and Identification Properties
Researchers developed the normal map-based proximal stochastic gradient method (NSGD), a novel algorithm that solves sto...
A Normal Map-Based Proximal Stochastic Gradient Method: Convergence and Identification Properties
The Normal Map-based Proximal Stochastic Gradient Method (NSGD) is a novel optimization algorithm that overcomes a key l...
A Global Optimization Algorithm for K-Center Clustering of One Billion Samples
Researchers have developed a novel global optimization algorithm for the K-center clustering problem, guaranteeing conve...
A Global Optimization Algorithm for K-Center Clustering of One Billion Samples
Researchers have developed a novel algorithm that guarantees finding the global optimum for the NP-hard K-center cluster...
A Global Optimization Algorithm for K-Center Clustering of One Billion Samples
A groundbreaking algorithm guarantees mathematically provable global optimality for the K-center clustering problem, han...
A Global Optimization Algorithm for K-Center Clustering of One Billion Samples
Researchers have developed the first practical global optimization algorithm for the K-center clustering problem, capabl...
Importance Weighting Correction of Regularized Least-Squares for Target Shift
A new theoretical analysis demonstrates that importance-weighted kernel ridge regression maintains optimal convergence r...
Importance Weighting Correction of Regularized Least-Squares for Target Shift
A theoretical study demonstrates that importance-weighted kernel ridge regression achieves minimax-optimal convergence r...
Importance Weighting Correction of Regularized Least-Squares for Target Shift
A theoretical analysis demonstrates that importance-weighted kernel ridge regression maintains optimal statistical perfo...
Importance Weighting Correction of Regularized Least-Squares for Target Shift
A theoretical analysis demonstrates that importance-weighted kernel ridge regression achieves optimal convergence rates ...
Network Topology Optimization via Deep Reinforcement Learning
A novel deep reinforcement learning algorithm called DRL-GS efficiently designs optimal network topologies by navigating...
Beyond State-Wise Mirror Descent: Offline Policy Optimization with Parameteric Policies
A theoretical breakthrough in offline reinforcement learning extends provable guarantees from finite to continuous actio...
SigmaQuant: Hardware-Aware Heterogeneous Quantization Method for Edge DNN Inference
SigmaQuant is an adaptive layer-wise heterogeneous quantization framework that intelligently assigns different bitwidths...
SigmaQuant: Hardware-Aware Heterogeneous Quantization Method for Edge DNN Inference
SigmaQuant is an adaptive layer-wise heterogeneous quantization framework designed for efficient deep neural network inf...
SigmaQuant: Hardware-Aware Heterogeneous Quantization Method for Edge DNN Inference
SigmaQuant is an adaptive layer-wise heterogeneous quantization framework designed for edge and mobile DNN deployment. I...