josh abbott
<p>Apr 16, 2016 06:56 PM</p>
<p>
sched. to give FriSem on 4/15 but cancelled.</p>
<p>
abstract is below: (looks important)</p>
<p>
<b>Good examples from categories and good categories from examples</b>...
Building intelligent machines that can make flexible and appropriate generalizations from limited evidence requires an understanding of human cognition. In this talk I will present two case studies which combine methods from cognitive science and machine learning to investigate how the mind reasons with abstract semantic knowledge. In particular, I will discuss how to extract good examples from categories and, conversely, <b>how to construct good categories from a few limited example</b> observations.</p>
<p>
In both studies I perform large-scale experiments to evaluate how people may reason about these ideas and investigate the challenges of scaling up formal automated models to perform similarly.</p>