Deep Reinforcement Learning For Optimal Order Placement In A Limit Order Book