Active Learning with SVMs

During the summer 2010, I visited for three months the Biocircuits Institute at UC San Diego for an internship under the supervision of Dr. Ramón Huerta.

I developed an active sampling algorithm for Support Vector Machines by defining an uncertainty measure based on the Platt’s approximation of the conditional probability to the SVMs. See poster.

I later used similar ideas to devise experiments in crystallization experiments in a Chemistry lab.