Approximate policy iteration with unsupervised feature learning based on manifold regularization | IEEE Conference Publication | IEEE Xplore