Composing Machine Learning via Market Mechanisms

Recent results have established striking and deep connections between prediction marketseconomic mechanisms that aggregate opinion about uncertain eventsand algorithms that can learn from expert advice. This project aims to extend these connections to further our understanding of the foundations and principles of markets and learning. The insights gained by doing so will be used to develop new methods for composing prediction technology, including machine learning, data mining algorithms, and human-driven mechanisms for prediction such as markets and crowdsourcing. These methods will enable the combination of these technologies to solve complex problems with distributed and otherwise incompatible sources of information.