Adaptive sample collection using active learning for kernel-based approximate policy iteration | IEEE Conference Publication | IEEE Xplore