The Cognitive Science Seminar Series welcomes guest presenter Shreya Karippurathu Rajagopal, who will give a talk titled "Towards Building a Comprehensive Computational Model of Classical Conditioning."

All cognitive science seminars will be virtual this semester. For Zoom passcode information, or to be added to the seminar distribution list, please email cogsci-seminar-requests@umich.edu.

ABSTRACT
Classical conditioning is a way of learning about one’s environment wherein a previously neutral stimulus (conditioned stimulus - CS) gets associated with another stimulus that evokes natural reward (unconditioned stimulus - US). Over several trials informing this pairing, the animal begins to generate a response (Conditioned response - CR) when only the CS is present. Classical Conditioning as a field encompasses a host of known phenomena about how the CR is affected by manipulating a range of factors in an experiment - the interstimulus interval, the number of CS-s, the order in which CS-s appear, etc. A key challenge in the field is to build a comprehensive computational model that can explain all these phenomena.

We approached this problem by familiarizing ourselves with some of the most prominent models of classical conditioning as of today so as to understand their shortcomings and build on them. The talk will go into the details of the temporal difference model of classical conditioning (Ludvig et. al 2012) and our attempts in replicating this model. It will also briefly touch upon other well-accepted models and our tentative plan for the project going forward.