stochastic-approximation

⚡ Single-pass algorithms for statistics

Code for Paper (Policy Optimization in RLHF: The Impact of Out-of-preference Data)

Stochastic Approximation Cut Algorithm for Inference in Modularised Bayesian Models

Code of the paper "Multi-Task Meta-Learning Modification with Stochastic Approximation".

Stochastic approximation with gym

Python implementations and some experiments for stochastic approximation algorithms

Published at TMLR in 2026. Focusing on the general stochastic approximation framework, it proposes a federated algorithm that finds the optimum of an average of contractive operators.

Synthetic threshold-payoff portfolio optimization on the probability simplex using projected CRN-SPSA and multistart simulation.

Self-calibrating control limits for streaming quality monitoring — a checkable stabilization threshold for adaptively preconditioned one-pass inference. Reproduction code.

A comprehensive research framework for Temporal Difference learning, implementing 9 state-of-the-art algorithms with theoretical analysis, empirical evaluation, and interactive visualization. Features custom environments, convergence proofs, bias-variance tradeoff analysis, and reproducible experimental pipelines for reinforcement learning research

Hands-on implementations of Reinforcement Learning algorithms from scratch, progressing from classical methods to deep RL with practical experiments.

A kernel-based stochastic approximation (KBSA) framework for contextual optimization.

One-pass Wald confidence intervals from averaged SA-Adam — preconditioning and momentum are asymptotically invisible. Reproduction code.

Published at IEEE Asilomar 2025. We study a general distributed heterogeneous stochastic approximation problem with M agents. The proposed DisSACC method converges to the desiderata with linear speedups, no heterogeneity bias, and near constant communication.