Improving Generalization of Reinforcement Learning Using a Bilinear Policy Network | IEEE Conference Publication | IEEE Xplore