Generalization Guarantees on Data-Driven Tuning of Gradient Descent with Langevin Updates

Develop a lightweight PyTorch implementation of Langevin Gradient Descent for convex regression tasks. Demonstrate that tuning hyperparameters via this method achieves optimal Bayes performance on standard benchmarks.

Suggested repo: LangevinLearn

"Hyperparameter tuning that mathematically guarantees Bayes optimality."

Estimated effort: 30h