Kavli Institute NeuroLunch
Tuesday, February 2nd, 2016
Kui Tang (PhD Candidate), Jebara Lab
"Bethe Learning of Graphical Models via MAP Decoding"
I am a second year PhD student in Computer Science at Columbia University.
Graphical models are a flexible and powerful method to model high-dimensional correlated data, including images and measurements of neural activity. Traditionally, these methods are computationally expensive. I will present a novel state of the art approximate algorithm for learning graphical models based on convex optimization principles that leveraged recently developed specialized linear programming solvers. I will conclude with applications from news and financial data and neuroscience. Joint work with Nicholas Ruozzi, David Belanger, and Prof. Tony Jebara.
Rm. 900 Sherman Fairchild Bldg. - Note the new room location
12:00 pm - 1:00pm
1212 Amsterdam Ave.
New York, NY 10027