What are the biophysical mechanisms underlying neuronal computations? Naively put, what are the diodes and the transistors underlying the computations at the level of individual neurons? Do cortical cells simply integrate over a large number of smallish inputs or can they perform more complex operations (such as coincidence detection)? Research carried out in our group as part of a program called Biophysics of Computation uses the detailed biophysics and microanatomy of cortical neurons to study their complexity from an information theoretical point of view.
We are interested in the biophysical origin of neuronal noise sources and in how they limit signal detection, signal estimation and the precision of spikes (Amit Manwani and Dr. Peter Steinmetz). Many of the biophysical components of a neuron have limitations which cause them to behave in a probabilistic and unreliable way (shown at right). These components thus act as noise sources which may interfere with the accurate transmission and computation.
All of this biophysics reseach has made its way into a monograph, entitled Biophysics of Computation: Information Processing in Single Neurons, authored by Prof. Christof Koch.
We continue to analyze the variability, reliability and randomness of neurons embedded in networks in order to understand the nature of the code used by nerve cells to transmit information. We work with various theoretical models (Poisson process; integrate-and-fire units) as well as with data from cortical pyramidal cells (Amit Manwani), neurons along the electro-sensory pathway of electric fish (Eigenmannia) (Dr. Fabrizio Gabbiani and Gabriel Kreiman; in collaboration with Prof. Walter Metzner at UC Riverside), cat LGN cells (Dr. Pamela Reinagel), and cells from various cortical areas in the monkey.
Can we derive unsupervised learning algorithms at the single neuron level (based on maximizing some information theoretical measure) that can start out with any set of voltage-dependent channels and end up with a neuron generating spikes (work done in collaboration with Dr. Martin Stemmler, now a post-doctoral fellow in Berlin, Germany).
The majority of synapses in the mammalian cortex originate from cortical neurons. Indeed, the largest input to cortical cells comes from neighbouring excitatory cells (about 90% of all excitatory synapses). However, most models of cortical development and processing do not reflect the anatomy and physiology of feedback excitation and are restricted to serial feedforward excitation. In collaboration with Professor Rodney Douglas in Zürich, our research foccuses on understanding the circuit properties of small cortical networks with massive feedback (primarily in visual cortex) based on the Canonical Microcircuit concept of Douglas and Martin (Gary Holt).