Event:

Cognitive tasks require the joint activity of a large population of neurons. Hence there is no a priori reason to find single neurons with easily interpretable activity profiles. Yet when we record from single neurons, we look for and bias our models by precisely these neurons. I will show the advantages of considering the “not easily interpretable” neurons recorded during a delayed vibrotactile discrimination task. The results of this analysis point to the importance of neural dynamics as a computational mechanism – a point that will be examined through the analysis of a noisy sensory discrimination task and a model of pitch detection. The approach taken in modeling all these tasks is one of training generic neural circuits without specifying a candidate mechanism – the merits and shortcomings of this approach will be discussed.