Reinforcement Learning: An On-Line Framework Using Support Vectors | IEEE Conference Publication | IEEE Xplore