Perceptual Science Series

Optimality and Probabilistic Computation in Visual Categorization

Dr. Wei Ji Ma

From Monday, February 17, 2014 -  11:00am
To Sunday, February 16, 2014 - 12:00am

New York University, Department of Psychology and Center for Neural Science

Copy to My Calendar (iCal) Download as iCal file

In models of perception under uncertainty, the terms “optimal” and “probabilistic” are often used interchangeably. I will argue that they have distinct meanings and are associated with different experimental designs. I will then present two recent studies: a) a complex visual search task with simple stimuli, in which optimality, not simplicity seems to govern behavior; b) a visual categorization task under ambiguity, in which human and monkey behavior is probabilistic, but  not necessarily optimal. In the latter task, I also show how the computation can be implemented by a plausible neural network, using a coding scheme known as probabilistic population coding. Finally, I will discuss V1 population recordings from monkeys performing this task.

Dr. Wei Ji Ma