I have just started working for Professor Aurel Lazar a few weeks ago, so this page will be updated more and more as my work progesses.
My current research interest is exploring spike-based algorithms for audio processing. One of the implications of the auditory system as filterbank is that a full sampling of an audio signal can be implemented in neurons with a firing rate far below Nyquist. Thus, I am working to build a system whereby I can perform processing and adjustments of audio signals in the spike domain, and then recover the audio signal in order to evaluate the effectiveness of the processing in the time/frequency domain. I am generating spikes using the simulated neuron of Lazar (2003-4) Time Encoding with an Integrate-and-Fire Neuron with a Refractory Period
Following that, how can we learn from spike trains? Are spikes a plausible representation of data in the neocortex? Is there some kind of method to the madness? E.g. can we design coritcal circuitry the way a VLSI designer lays out transistors?
It seems that for any complicated system to work properly, it is imperative to use levels of abstraction. The idea is that design at each level can done according to specifications, while providing service to the level above and below. For example, in digital circuits, logic designers provide specifications to layout designers without regard to design rules, while layout level designers use the transistor diagram as their guide to laying out semiconductors and doped channels. Similarly, cortical unit "designers" (e.g. darwin) can abstract away the physiology of neurons and only work with a functional description. A rough sketch of wthe parallels between levels of abstraction in computer design with VHDL and brain organization is as follows:
Obviously, in order to be a good engineer you need to know a good deal about ALL levels in the process...