←── back to feed
/topics/arxiv-machine-learning-papers-october-9

arXiv machine learning papers October 9

50 items●1 sources●updated 30d ago●trend 0

On October 9, 2026, arXiv published 20 machine learning papers spanning theoretical foundations, sampling methods, and practical applications. Topics ranged from information-theoretic analyses of VC learning and kernel bandits to new techniques for conformal prediction, synthetic data generation with causal fidelity, and robust system identification under adversarial corruption.

  • eCMI framework recovers optimal PAC guarantees for VC classes through algorithm-dependent information-theoretic analysis
  • JevForest achieves 0.729 accuracy on AG News with four-question feature acquisition budget
  • Conformal prediction under partial verification matches complete verification with less than 2× minimum verification cost
  • σTransfer enables zero-shot uncertainty transfer from small to large neural networks under μP scaling
  • NALD and NRALD extend non-reversible anchored Langevin dynamics to non-smooth, heavy-tailed sampling on constrained domains
[BLG]blog/rss50
The optimal information complexity of VC learning
arXiv stat.ML · Steve Hanneke, Juexiao Wang · 30d
JevForest: Path Voting for Budgeted Feature Acquisition
arXiv stat.ML · Yu Yan · 30d
From Log-Odds to Shapley Values: An Explanatory Geometry for the Weighted Naive Bayes Classifier
arXiv stat.ML · Vincent Lemaire, Fabrice Cl\'erot · 30d
What can linear attention learn from nonlinear teachers in-context?
arXiv stat.ML · Mary Letey, Arman Rysmakhanov, Yue M. Lu, Cengiz Pehlevan, Jacob Zavatone-Veth · 30d
Calibrating Ambiguity Set via Diagnostic Transport for Distributionally Robust Optimization
arXiv stat.ML · Wenbin Zhou, Elizabeth Cucuzzella, Shixiang Zhu · 30d
Conformal Prediction under Partial Verification
arXiv stat.ML · Zijun Yu, Yu Gu, Vahid Partovi Nia, Masoud Asgharian · 30d
Transformed Samplers with Variance Reduction
arXiv stat.ML · Siran Liu, Michalis Tisias, Petros Dellaportas · 30d
A General $\widetilde{\Omega}(\sqrt{T \gamma_T})$ Lower Bound for Kernel Bandits
arXiv stat.ML · Chenkai Ma, Jonathan Scarlett · 30d
Accelerating Non-Smooth and Heavy-Tailed Sampling
arXiv stat.ML · Pervez Ali, Xiaoyu Wang, Yingli Wang, Lingjiong Zhu · 30d
Beyond Distributional Fidelity: Causal-Penalized Diffusion for Synthetic Tabular Data
arXiv stat.ML · Lan Tao, Yongxian He, Shirong Xu, Yidong Ouyang, Guang Cheng · 30d
Feature Space Adaptation for Effortless Gaussian Process Flows
arXiv stat.ML · Thomas Cowperthwaite, Louis Sharrock, Lachlan Astfalck, Henry Moss · 30d
PSI-SINDy: Post-Selection Inference for Sparse Identification of Nonlinear Dynamics
arXiv stat.ML · Ashraful Islam, Shuichi Nishino, Tomohiro Shiraishi, Ichiro Takeuchi · 30d
LAIR-Net: Leaky Alignment-Impulse Residual Networks for Tabular Regression
arXiv stat.ML · Rahul Goswami, Aryan Bhambu, Bittu Karmakar · 30d
Embedding-Bias in Conditional Independence Testing
arXiv stat.ML · Nikolaj Thams, Anton Rask Lundborg · 30d
Minimax Gaussian Mechanisms for Continual Machine Unlearning
arXiv stat.ML · Qi Kuang, Yin Xia · 30d
$\sigma$Transfer: Uncertainty Transfer from Small to Large Networks under $\mu\mathrm{P}$
arXiv stat.ML · Richard Bergna (University of Cambridge, Spotify), Fernando Ruiz Mazo (University of Cambridge), Nicol\`o Felicioni (Spotify), Jos\'e Miguel Hern\'andez-Lobato (University of Cambridge), Kamil Ciosek (Spotify) · 30d
Softmax Attention on Gaussian Mixtures: Linear When It Can, Selective When It Must
arXiv stat.ML · Simon Gabet (LMO), Etienne Boursier (LMO, CELESTE), Claire Boyer (LMO, IUF) · 30d
Conditional Kernel Stein Discrepancy
arXiv stat.ML · Federico Matteucci, Florian Kalinke · 30d
Learning structured linear dynamical systems from missing observations
arXiv stat.ML · Aravinda Kanchana Ruwanpathirana, Hemant Tyagi, Sunny G. W. Wang · 30d
RobustLDS: Learning linear dynamical systems under adversarial corruptions
arXiv stat.ML · Aravinda Kanchana Ruwanpathirana, Hemant Tyagi · 30d
Score-Based Learning of Cluster DAGs from Interventions
arXiv stat.ML · Gaetano Tedesco (University of Amsterdam), Alex Markham (University of Copenhagen) · 30d
Efficient quadratic entropy with distance sketches
arXiv stat.ML · Steve Huntsman · 30d
Efficient and Generalizable Archetypal Analysis for Discrete Data
arXiv stat.ML · A. Emilie J. Wedenborg, Jesper L{\o}ve Hinrich, Morten M{\o}rup · 30d
Diffusion Removes Langevin's Conditioning Dependence: A Sharp Gaussian Analysis
arXiv stat.ML · Adam Perbost, Francis Bach, Pierre Marion · 30d
Differentiable Systematic Resampling for Variational Sequential Monte Carlo
arXiv stat.ML · Fredrik Cumlin, Saikat Chatterjee · 30d
Quickest Change Detection with Diffusion-Integrated Scores
arXiv stat.ML · Arman Adibi, Mohammadreza Maleki, Sanjeev Kulkarni, H. Vincent Poor · 30d
ISBO: Scalable Spatio-Temporal Bayesian Optimization with Log Gaussian Cox Process Models via the INLA-SPDE Approach
arXiv stat.ML · Kaichuang Yang, H{\aa}vard Rue, Jakob Zeitler · 30d
Testing Algebraic Complete Intersections
arXiv stat.ML · Alessandro Tamai · 30d
Prediction-Powered Data Fusion for Treatment Effect Estimation
arXiv stat.ML · Yonghan Jung, Shu Yang · 30d
Density Ratio Estimation with Stein Displacement Fields
arXiv stat.ML · Song Liu · 30d
Discovering Global False Negatives On the Fly for Self-supervised Contrastive Learning
arXiv stat.ML · Vicente Balmaseda, Bokun Wang, Ching-Long Lin, Tianbao Yang · 30d
Exact SO(3)-Equivariant Isotropic Kernels for Rotation-Robust Neural Dynamics
arXiv stat.ML · Ridham Patel · 30d
D-SLR: The Disjoint Row-Sparse plus Low-Rank Decomposition
arXiv stat.ML · Vincent Szolnoky · 30d
How Many Directions Must a Truncated Diffusion Sampler Retain? Matching Bounds Under Power-Law Spectra
arXiv stat.ML · Radmehr Karimian, Ali Mohades, Johannes Lederer · 30d
Explaining the Saliency Map Sparsity of Adversarially-Trained Neural Networks
arXiv stat.ML · Yannick Lunk, Atell Yehor Krasnopolsky, Damien Garreau, Leon Bungert · 30d
SPD-MetaFormer is what you need for small-data brain decoding
arXiv stat.ML · Zhida Wang, Wei Lyu, Guo Yu, Sui Tang · 30d
Decision-Sufficient Posterior Approximation
arXiv stat.ML · Sean Plummer · 30d
Tight Bounds for Equivalence Testing with Non-Adaptive Conditional Samples
arXiv stat.ML · Gautam Kamath · 30d
When Lower Reconstruction Loss Hurts: Distributionally Robust Refinement for Low-Bit LLM Quantization
arXiv stat.ML · Yanlong Zhao, Xiaoyuan Cheng, Huihang Liu, Baihua He, Xinyu Zhang, Harrison Bo Hua Zhu, Wenlong Chen, Li Zeng, Zhuo Sun · 30d
Regularized Small Area Estimation with Graph Laplacian Benchmarking priors
arXiv stat.ML · Jairo F\'uquene-Pati\~no, Brenda Betancourt · 30d
From Geometry to Generalization: Why Row Normalization Can Beat Adam and Muon
arXiv stat.ML · Jihwan Kim, Dogyoon Song, Chulhee Yun · 30d
Sequential Conditional Independence Testing with Machine Learning Models
arXiv stat.ML · Angel Reyero-Lobo, Michele Meziu, Sebastian Uriel Arias, Peter Gr\"unwald · 30d
Rare Gate Disagreements Can Limit Plasticity: When Gradient Flow Mispredicts Finite-Batch SGD
arXiv stat.ML · Ruoyu Zhao, Mingxuan Zhang, Jianbo Dai, Jiaqi Wu, Chenyu Zhu, Tong Che · 30d
Automated Detection of Match Phases in Football from Spatio-Temporal Tracking Data Using Graph Neural Networks
arXiv stat.ML · Vincent Renner, Nils Koster, Pascal Bauer, Melanie Schienle · 30d
Optimal random quantisers for spherically symmetric distributions
arXiv stat.ML · Luc Pronzato, Anatoly Zhigljavsky · 30d
Exploiting Gradients in Bayesian Inference of Expensive Simulators
arXiv stat.ML · \v{S}imon Sold\'at, V\'aclav \v{S}m\'idl · 30d
Credal Machine Learning for Risk-Averse Decision Making
arXiv stat.ML · Timo L\"ohr, Paul Hofman, Maximilian Muschalik, Eyke H\"ullermeier · 30d
Verification with Transfer: Exact Information Frontiers and Their Price in Calls
arXiv stat.ML · Hazar Yueksel · 30d
Composite Online-to-Nonconvex Conversion with Optimal Oracle Complexity
arXiv stat.ML · Mingyi Li, Taira Tsuchiya, Kenji Yamanishi · 30d
Closing the Horizon Gap in Policy Optimization for Adversarial MDPs
arXiv stat.ML · Mingyi Li, Taira Tsuchiya · 30d