AI Research News Feeds for September 28th, 2025

AI RESEARCH PAPERS & ACADEMIC SOURCES

• EF21 with Bells & Whistles: Six Algorithmic Extensions of Modern Error Feedback
• Actor-Critic learning for mean-field control in continuous time
• Bagged k-Distance for Mode-Based Clustering Using the Probability of Localized Level Sets
• Laplace Meets Moreau: Smooth Approximation to Infimal Convolutions Using Laplace's Method
• On Adaptive Stochastic Optimization for Streaming Data: A Newton's Method with O(dN) Operations
• Dynamic Bayesian Learning for Spatiotemporal Mechanistic Models
• On the Ability of Deep Networks to Learn Symmetries from Data: A Neural Kernel Theory
• Fine-grained Analysis and Faster Algorithms for Iteratively Solving Linear Systems
• Deep Generative Models: Complexity, Dimensionality, and Approximation
• ClimSim-Online: A Large Multi-Scale Dataset and Framework for Hybrid Physics-ML Climate Emulation
• Conditional Wasserstein Distances with Applications in Bayesian OT Flow Matching
• Deep Variational Multivariate Information Bottleneck - A Framework for Variational Losses
• Diffeomorphism-based feature learning using Poincaré inequalities on augmented input space
• Finite Expression Method for Solving High-Dimensional Partial Differential Equations
• Randomly Projected Convex Clustering Model: Motivation, Realization, and Cluster Recovery Guarantees
• Minimax Optimal Deep Neural Network Classifiers Under Smooth Decision Boundary
• Optimal and Efficient Algorithms for Decentralized Online Convex Optimization
• Characterizing Dynamical Stability of Stochastic Gradient Descent in Overparameterized Learning
• PREMAP: A Unifying PREiMage APproximation Framework for Neural Networks
• Score-Aware Policy-Gradient and Performance Guarantees using Local Lyapunov Stability
• On the O(sqrt(d)/T^(1/4)) Convergence Rate of RMSProp and Its Momentum Extension Measured by l_1 Norm
• Categorical Semantics of Compositional Reinforcement Learning
• Transformers from Diffusion: A Unified Framework for Neural Message Passing
• Optimal Sample Selection Through Uncertainty Estimation and Its Application in Deep Learning
• Actor-Critic learning for mean-field control in continuous time
• Modelling Populations of Interaction Networks via Distance Metrics
• BitNet: 1-bit Pre-training for Large Language Models
• Physics-informed Kernel Learning
• Last-iterate Convergence of Shuffling Momentum Gradient Method under the Kurdyka-Lojasiewicz Inequality
• Posterior and Variational Inference for Deep Neural Networks with Heavy-Tailed Weights
• Maximum Causal Entropy IRL in Mean-Field Games and GNEP Framework for Forward RL
• Degree of Interference: A General Framework For Causal Inference Under Interference
• Quantifying the Effectiveness of Linear Preconditioning in Markov Chain Monte Carlo
• Sparse SVM with Hard-Margin Loss: a Newton-Augmented Lagrangian Method in Reduced Dimensions
• On Model Identification and Out-of-Sample Prediction of PCR with Applications to Synthetic Controls
• Bayesian Scalar-on-Image Regression with a Spatially Varying Single-layer Neural Network Prior
• Linear Separation Capacity of Self-Supervised Representation Learning
• On the Convergence of Projected Policy Gradient for Any Constant Step Sizes
• Learning with Linear Function Approximations in Mean-Field Control
• A New Random Reshuffling Method for Nonsmooth Nonconvex Finite-sum Optimization
• Model-free Change-Point Detection Using AUC of a Classifier
• EF21 with Bells & Whistles: Six Algorithmic Extensions of Modern Error Feedback
• Multiple Instance Verification
• Learning from Similar Linear Representations: Adaptivity, Minimaxity, and Robustness
• Exponential Family Graphical Models: Correlated Replicates and Unmeasured Confounders, with Applications to fMRI Data
• Optimizing Return Distributions with Distributional Dynamic Programming
• Imprecise Multi-Armed Bandits: Representing Irreducible Uncertainty as a Zero-Sum Game
• Early Alignment in Two-Layer Networks Training is a Two-Edged Sword
• Hierarchical Decision Making Based on Structural Information Principles
• Generative Adversarial Networks: Dynamics
• “What is Different Between These Datasets?” A Framework for Explaining Data Distribution Shifts
• Assumption-lean and data-adaptive post-prediction inference
• Bagged Regularized k-Distances for Anomaly Detection
• Four Axiomatic Characterizations of the Integrated Gradients Attribution Method
• Fast Algorithm for Constrained Linear Inverse Problems
• High-Rank Irreducible Cartesian Tensor Decomposition and Bases of Equivariant Spaces
• Best Linear Unbiased Estimate from Privatized Contingency Tables
• Interpretable Global Minima of Deep ReLU Neural Networks on Sequentially Separable Data
• Enhanced Feature Learning via Regularisation: Integrating Neural Networks and Kernel Methods
• Data-Driven Performance Guarantees for Classical and Learned Optimizers
• Contextual Bandits with Stage-wise Constraints
• Boosting Causal Additive Models
• Frequentist Guarantees of Distributed (Non)-Bayesian Inference
• Asymptotic Inference for Multi-Stage Stationary Treatment Policy with Variable Selection
• EMaP: Explainable AI with Manifold-based Perturbations
• Autoencoders in Function Space
• Nonparametric Regression on Random Geometric Graphs Sampled from Submanifolds
• System Neural Diversity: Measuring Behavioral Heterogeneity in Multi-Agent Learning
• Distribution Estimation under the Infinity Norm
• Extending Temperature Scaling with Homogenizing Maps
• Density Estimation Using the Perceptron
• Simplex Constrained Sparse Optimization via Tail Screening
• Score-Based Diffusion Models in Function Space
• Regularized Rényi Divergence Minimization through Bregman Proximal Gradient Algorithms
• WEFE: A Python Library for Measuring and Mitigating Bias in Word Embeddings
• Frontiers to the learning of nonparametric hidden Markov models
• On Non-asymptotic Theory of Recurrent Neural Networks in Temporal Point Processes
• Classification in the high dimensional Anisotropic mixture framework: A new take on Robust Interpolation
• Universal Online Convex Optimization Meets Second-order Bounds
• Sample Complexity of the Linear Quadratic Regulator: A Reinforcement Learning Lens
• Randomization Can Reduce Both Bias and Variance: A Case Study in Random Forests
• skglm: Improving scikit-learn for Regularized Generalized Linear Models
• Losing Momentum in Continuous-time Stochastic Optimisation
• Latent Process Models for Functional Network Data