A memory-based reinforcement learning algorithm for partially observable Markovian decision processes | IEEE Conference Publication | IEEE Xplore