Can Serif Mekik, RPI Graduate Student

 

Can Serif Mekik, RPI Graduate Student

Sage 4101

April 26, 2017 12:00 PM - 1:30 PM

ABSTRACT:

Raven's Matrices are a family of intelligence tests sharing a rather elegant format. These tests are well-suited to the study of general intelligence, mental representations, and how different mental (perceptual, cognitive, motivational, etc.) processes interact, and they attract an increasing amount of attention from the cognitive modeling and artificial intelligence communities on account of their simple format and broad relevance.

 

In this talk, I discuss plans to develop a model which integrates empirical insights about the role of perceptual and motivational processes in affecting human performance on Raven's matrices, and I report on my current progress toward this goal. In particular, I present a preliminary model which aims to simulate perceptual and cognitive processes implicated in some Raven's style problems by integrating  deep learning (convolutional neural network) with a rule system. I argue that the model can be further developed to account for many cognitive and motivational issues.

 

Add to calendar
Share|