Probabilistic Programming for Algorithmic Trading
Probabilistic programming is often advertized as “democratizing machine learning”, because it provides a very simple but expressive model of statistical inference. The purpose of this thesis is to validate that claim by applying it to algorithmic trading of securities. More specifically, we want to apply it for “auto-tuning” of parameters for trading strategies. In the thesis, no actual trading will take place, but we will evaluate the approach with historical ticker data.