Showing math.OCShow all
3 papers · 1 filter
math.OC2024
A Large Deviations Perspective on Policy Gradient Algorithms
Wouter Jongeneel, Daniel Kuhn, Mengmeng Li
Motivated by policy gradient methods in the context of reinforcement learning, we identify a large deviation rate function for the iterates generated by stochastic gradient descent…
math.OC2024
On continuation and convex Lyapunov functions
Wouter Jongeneel, Roland Schwan
Suppose that the origin is globally asymptotically stable under a set of continuous vector fields on Euclidean space and suppose that all those vector fields come equipped with --…
math.OC2024
Small errors in random zeroth-order optimization are imaginary
Wouter Jongeneel, Man-Chung Yue, Daniel Kuhn
Most zeroth-order optimization algorithms mimic a first-order algorithm but replace the gradient of the objective function with some gradient estimator that can be computed from a…