Machines can learn to perform certain recognition and prediction tasks, meeting or exceeding human-level performance, but require far more training examples than humans need. MadLab researchers are developing new ML theory and algorithms that can learn faster and with far less supervision.
