The NAC approach

Neural assembly computing (NAC) is an approach for investigating and exploring the computation performed by ‘neural cell assemblies’.

We are concerned with the way assemblies interact and how such interactions result in information processing, involving causal and hierarchical relations. We investigate how assemblies represent information, how they control the data flux carried by spike streaming, how they create parallel processes by branching and dismantling other coalitions, how they reverberate and create memory, among other issues.

We call digital assemblies a set of neurons operating in two stable states: all neurons in an assembly is firing during a time interval (the coalition is ON), or all members are silent (the assembly is OFF). Operating in the intermediate states, analog assemblies may change the number of firing members during a time window, the duration of their activity, or both the number of involved members and the assembly’s active lifetime.

In addition to the synaptic weights, the spike propagation delay plays an important role in NAC. The neural assembly approach is implemented in spiking neural networks.