Proposes a Gaussian Process-based computational graph completion framework, enabling robust inference of unknown functions and variables in complex systems.

Houman Owhadi

This paper analyzes adversarial attacks on Gaussian process bandits, proposing white-box and black-box methods that successfully steer algorithms toward target regions with low budgets.

Eric Han, Jonathan Scarlett

Proposes generalized kernel thinning (TARGET KT) with dimension-free error bounds, improving high-dimensional distribution compression.

Raaz Dwivedi, Lester Mackey

Moser Flow (MF), a divergence-based continuous normalizing flow, avoids ODE solving, enabling efficient sampling and density estimation on complex manifolds.

Noam Rozen, Aditya Grover, Maximilian Nickel et al.

Adaptive embedding in kernel and neural bandits achieves sample complexity depending only on effective dimension.

Yinglun Zhu, Dongruo Zhou, Ruoxi Jiang et al.

The paper introduces the Optimum-statistical Collaboration framework and VHCT algorithm for efficient black-box optimization.

Wenjie Li, Chi-Hua Wang, Guang Cheng et al.

Proposes IXOMD, a model-free algorithm for high-probability convergence to NE in two-player zero-sum partially observable Markov games, with rate O(1/√T).

Tadashi Kozuno, Pierre Ménard, Rémi Munos et al.

Using measure representation and sparse regularization, proves gradient descent can recover teacher network parameters with high probability in two-layer ReLU models.

Shunta Akiyama, Taiji Suzuki

Kernel Thinning compresses n points to √n with provably comparable MMD error, outperforming i.i.d. sampling.

Raaz Dwivedi, Lester Mackey

Proposes dataset inference leveraging model memorization to verify ownership with over 99% confidence, using statistical tests and distance estimation techniques.

Pratyush Maini, Mohammad Yaghini, Nicolas Papernot

Proposes SDE-BNN, an infinite-depth Bayesian neural network using stochastic differential equations, with a zero-variance gradient estimator for scalable, expressive inference.

Winnie Xu, Ricky T. Q. Chen, Xuechen Li et al.

Proposes TAX4CS framework integrating SHAP, LIME, and PDP to enhance transparency and auditability of credit scoring models.

Michael Bücker, Gero Szepannek, Alicja Gosiewska et al.

Proposes Linear Optimal Transport (LOT) embedding for high-dimensional distribution classification, nearly matching Wasserstein-2 distances, especially effective under shift and scale perturbations.

Caroline Moosmüller, Alexander Cloninger

Proposed algorithm-independent lower bounds for Gaussian process bandit optimization, improving error probability dependence in standard and robust settings.

Xu Cai, Jonathan Scarlett

Proposed low-rank tensor bandit algorithms with finite regret bounds, outperforming existing methods in high-dimensional online decision tasks.

Jie Zhou, Botao Hao, Zheng Wen et al.

Proposes that deep discriminative models are linearly identifiable in function space, validated on GPT-2, BERT, and simulated data, using nonlinear ICA theory.

Geoffrey Roeder, Luke Metz, Diederik P. Kingma

Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks.

Abdulkadir Canatar, Blake Bordelon, Cengiz Pehlevan

Extends NNGP and NTK to multi-head attention, proving Gaussianity in the infinite width limit with kernel expressions.

Jiri Hron, Yasaman Bahri, Jascha Sohl-Dickstein et al.

PatientBandits algorithm excels in handling stochastic bandits with arm-dependent delays.

Anne Gael Manegueu, Claire Vernade, Alexandra Carpentier et al.

Double GAN-based conditional independence test controls Type I error and detects dependencies effectively in high-dimensional data.

Chengchun Shi, Tianlin Xu, Wicher Bergsma et al.