Linear Stochastic Approximation: How Far Does Constant Step-Size and Iterate Averaging Go?

Chandrashekar Lakshminarayanan Csaba Szepesvári

Published in: AISTATS (2018)

Keyphrases

step size
stochastic approximation
steepest descent method
cost function
convergence rate
monte carlo
convergence speed
temporal difference
policy iteration
machine learning
reinforcement learning
temporal difference learning
multiscale
variable step size
learning rate
wavelet coefficients
neural network