Temporal Codes in Neuromorphic aVLSI



It is an ongoing discussion in the field of Neuroscience, how information is coded in the nervous system. Tradionally the spiking frequency of neurons were considerd to carry all relevant information from one neuron to another. But lately it was argued that the temporal patterns of spiketrains were relevant to the information processing in the nervous system too. In some parts of the nervous system of barn owls and chickens it is actually known, how the timing of spikes does convey information, namely in coding the direction in the horizontal plain of a soundsource. And some visual processing tasks are performed by humans in such a short time, that every neuron along the processing path can only emit a few nerve-pulses (action potentials, APs), so that speaking of an average frequency would not make sense (Mainly argued by Simon Thorpe).

So there are some clues, THAT temporal information is used in the nervous system, but there are not many general ideas around, as to HOW it is used. This project tries to work out ways of using temporal codes in practical applications, which some day might also result in insight in the ways in which the brain uses temporal codes.

Project Outline

The project's aim is to build neuromorphic aVLSI devices that take advantage of temporal encoding. On a scientific ground, the resulting algorithms may give us a clue on how the brain uses temporal codes. From an engeneering perspective neuromorphic analog VLSI (aVLSI) devices can often solve problems much more energy-efficient (less current, smaller) than more traditional approaches based on digital logic.
last update: 13.1.2003