Darwinian Algorithms in the Brain
- Speaker: Dr Chrisantha Fernando, Department of Informatics, Sussex University
- Date: Tuesday, 28 June 2011 from 16:45 to 17:45
- Location: Room 745, Malet Street
Can good ideas literally evolve in the brain overnight? It is known that copying of information takes place in the brain, for example, orientation selective receptive fields are copied from neurons in one part of visual cortex to adjacent neurons. For such entities to be true Darwinian replicators there must be covariance between their traits and their fitness as defined by Price. However, synaptic receptive fields are limited heredity replicators. We ask, is it possible for groups of receptive fields to be units of evolution? Gerald Edelman argued that neuronal groups could be Darwinian individuals, an argument with which Francis Crick strongly disagreed. Crick called Edelman's theory Neural Edelmanism rather than Neural Darwinism for it lacked the ability to copy functional traits that were dependent on patterns of synaptic connections from group to group. In other words, Edelman's neuronal groups are merely implementing competitive learning, rather than natural selection which requires transmission of information between individuals. Edelman's neuronal groups are neither units of evolution as defined by John Maynard-Smith nor Darwinian units as defined by Price. In an attempt to debug Edelman's theory and to take Francis Crick seriously, I describe a mechanism by which higher-order units of evolution could indeed replicate in the brain, using causal inference mediated by spike-time dependent plasticity (STDP) to copy microcircuits (patterns of synaptic connections) from one brain region to another. Furthermore, I propose a host of other higher-order neuronal entities that may plausibly exist and be interpreted as undergoing Darwinian evolution, such as paths of activity through a network. Such paths can be interpreted as embedded and overlapping units of evolution within a network. For the first time this allows the insights of population genetics to directly apply to cellular and systems neuroscience, creating a new field we could call Population Synaptics.