UCL Logo

Saliency, Attention, and Inference with Bayesian Neurons

Timm Lochmann

ENS, France

We illustrate how effects related to saliency and attention can be integrated in a model assuming neurons to be simple inference devices (Deneve, 2005). This provides a functional interpretation of why firing rates are multiplicatively scaled by changes in stimulus contrast or attentional state. We explore the impact of such scaling on the coincidence interval in the model and implications for detecting changes in the input regime.