"A forward oriented reinforcement learning approach to the problem of optimized trade execution"
This master's thesis tackles the important problem of optimized trade execution, which aims to obtain the best possible price for a trade instructed by higher level investment decisions. In its simplest form the problem is defined by a particular financial instrument, which must be bought or sold within a fixed time horizon, while minimizing the expenditure for doing so. Modifications to an existing reinforcement learning approach are made and a novel forward-learning algorithm is proposed, outperforming the cost avoiding capabilites of all previous methods.