A benchmark to test decision-making algorithms for contextual-bandits. The library implements a variety of algorithms (many of them based on approximate Bayesian Neural Networks and Thompson sampling), and a number of real and syntethic data problems exhibiting a diverse set of properties.
