Randomized Transport Maps for Model-Free Policy-Gradient Mean-Field Control
Transport REINFORCE is a model-free policy gradient method for Mean-Field Control that addresses standard REINFORCE's failure to account for how a policy affects the population distribution. It introduces transport maps to perturb the population law, allowing it to estimate this missing mean-field contribution. This results in a more accurate gradient estimator and superior performance on various control benchmarks compared to standard methods.