Context-dependent meta-control for reinforcement learning using a Dirichlet process Gaussian mixture model | IEEE Conference Publication | IEEE Xplore