Connecting Computer Science and Statistics Methods in
Temporal Data Mining
(Joint Seminar with ECE)
Dr. K.P. Unnikrishnan
General Motors Research
Place: 2250 Engineering
Abstract: Discovering frequent episodes from event streams has applications in areas ranging from automotive manufacturing to bio-informatics and neurobiology. We describe efficient algorithms for frequent episode discovery when the events have durations. We then connect these counting-type methods in Computer Science with Hidden Markov Models (HMMs) in Statistics. This allows us to determine the statistical significance of the discovered frequent episodes. We show use of these methods for throughput improvement and root-cause analysis in automotive assembly plants. We also illustrate their use for analyzing multi-neuronal data.
Biography: Dr K.P. Unnikrishnan received the PhD degree in Physics (biophysics)