Scheme: Wolfson Research Merit Awards
Organisation: Imperial College London
Dates: Apr 2010-Sep 2012
Summary: The brain is made of billions of neurons, which together form the world's most powerful information-processing machine. Somehow, these cells organize themselves into networks that coherently guide animal behavior. The idea of building learning machines along similar principles to the brain is an old one. However, despite several times when this goal seemed within reach, today’s artificial neural networks show nothing like the flexibility of the biological brain. This may indicate that there are biological processes involved in organizing neurons that are not captured in today’s theories of computational neuroscience.
This proposal aims to test a new theory for how the brain self-organizes at a large-scale level. It is based on the hypothesis is that strengthening of a neuron's output synapses stabilizes recent changes in the same neuron's inputs. If correct, this would enable neurons to self-organize through competitive interactions, analogously to a market economy. One can think of a neuron as like an entrepreneur, using "raw materials" (the information in its inputs) to manufacture a "product" (its output). When the neuron changes its input synapses, it changes the mix of raw materials it uses, thus producing a new product. If downstream neurons "purchase" this product by listening (i.e. strengthening their own input synapses), manufacture will continue. If not, the product will be discontinued. In the economy, interactions like this allow autonomous agents (people and firms) to organize into networks. Could similar interactions organize the billions of cells in the brain into a single coherent system? We will test this with a combination of computational network analysis, and experimental neurobiology.