We will replace the backward propagation part on the preceding neural network with a Monte Carlo algorithm, called the cross-entropy method. This is a general-purpose algorithm introduced by Reuven Rubinstein which is quite helpful in many cases, especially for rare event simulation. It has been proven efficient for many reinforcement learning tasks, so why not give it a try?
The method consists of two parts:
- Generate a random data sample (trajectories, vectors) according to a specified mechanism.
- Update the parameters of the random mechanism based on the data to produce a better sample in the next iteration. This step involves minimizing the cross-entropy or Kullback–Leibler divergence. ...