My lab works on the problem of learning to recognize, categorize and generate structural patterns based on examples (this is called induction). Structural patterns refers to patterns in data that are not easily represented by a fixed-length sequence of numbers (i.e. a vector). Examples: English text, the shape of a molecule, a sequence of DNA, a radio signal, a system log, a video. The lab is problem-driven (not solution driven), so we use a number of methods and tools like artificial neural networks, support vector machine, deep belief networks, hidden Markov models, evolutionary algorithms and deep learning. Problems are drawn from a variety of domains, but problems related to biology are especially welcome.