Efficient Gradient Estimation for Motor Control Learning

Gregory Lawrence, Noah Cowan, Stuart Russell

The task of estimating the gradient of a function in the presence of noise is central to several forms of reinforcement learning, including policy search methods. We present two te…