Adaptation and recurrent networks

I did my PhD at the Center for Molecular and Behavioral Neuroscience of Rutgers University Newark, in Prof. Bart Krekelberg’s lab.

Recurrent network model

The first part of my PhD consisted in implementing a well studied model of primary visual cortex and testing what happened if I “presented” the model with two oriented gratings in very quick succession, like so:

It turned out that the model exhibited a behavioral phenomenon known as adaptation: when the response to the second stimulus is contaminated by the response to the first, resulting in a different perceived orientation. Historically this was always attributed to plasticity, or changes within the synaptic weights, but my model displayed the same effect with no changes to weights, only as a consequence of the recurrent connections introducing a time lag to the response. This work got published in Cell Reports, and I wrote a plain language explanation of the results as well.

Effects on behaviour

I then tested the mathematical model in the real world using psychophysics tests, such as showing people images in quick succession and asking what they see. I found that the perceptual effects of showing these two oriented gratings in close succession, and for a short period of time, is consistent with the model predictions, further validating the idea! These results were published in Frontiers in Systems Neuroscience.