Resume in html
My advisor is Eric Mjolsness of the Machine Learning Systems group at JPL. I am also a part of the GURU research group at UCSD.
My thesis project is the Fluid Feature Hierarchy, a specialization for vision of the Stochastic Grammar framework developed by Eric. A central application of my thesis will be the recognition and classification of craters in planetary images, done in collaboration with the MLS group at JPL.
Part of the FFH research program involves the study of how to allocate limited processing resources to the most salient aspects of the problem, a form of attention. Here is a simple (alpha quality) demo of a standard Hopfield net with attention included. The net successively relaxes a group of units that it judges most salient, using a theoretical result for Hopfield/Grossberg dynamics established in previous research by Eric.
This is my thesis proposal for the FFH
project.
Email me at lramsey@ucsd.edu