Heterogeneity, non-linearity and dimensionality: how neuron and network properties shape computation

The brain is a unique physical system, in that its dynamics have a clear function: making its owner respond to the world around it. The physical structure of the brain (its ‘hardware’) shapes this information processing and vice versa: the computations needed for information processing (the ‘software’) are adapted to the physical structure of the hardware. Moreover, the brain has the unique property that it is able to adapt its hardware to the task at hand. I will discuss a few of the most striking differences between real neural structures and their commonly used artificial counterparts, and the consequences for different types of computation. I will show how neural properties (non-linearities) differ from neuron to neuron (heterogeneity), and their consequences for network activity (dimensionality) and computation.

Workshop main page