Temporal feature analysis in brain-inspired neural systems

Abstract

The brain identifies potentially salient features within continuous information streams, but the underlying mechanisms are poorly understood. I will show two biologically inspired neural network models that perform such analyses. The seemingly different models suggest a common principle, which we term self-consistent mismatch detection, for temporal feature analyses. A network of two-compartment neuron model implementing this principle conducts a surprisingly wide variety of temporal feature analysis. The model is also potentially useful in neural engineering.