Dimension-Uniform Discretization Analysis of Preconditioned Annealed Langevin Dynamics for Multimodal Gaussian Mixtures

Collection of recent machine-learning theory and methodology papers covering score-based/diffusion samplers, divergence estimation, calibration for survival models, causal robustness, hypergraph generative models, Fourier analyses of learning dynamics, distribution-valued time-series forecasting, operator learning for SPDEs, and robustness-adaptive multi-task and matrix-completion methods. Highlights include (1) dimension-uniform discretization analysis for preconditioned annealed Langevin dynamics with an exponential-integrator stability result; (2) StAD, a Langevin–Stein based method to dist