Synaptic plasticity: From single neuron to cortical network

Visual information arises from the retinae of an animal and is passed to the cortical cells via the primary visual pathway. Most simple cells in the cortex respond to orientation features in the environment. A layer of simple cells forms an orientation map that contains orientation vortex and patches. In order to explain these experimental facts, synaptic modification rules based on the second order statistics of the environment (correlation-based learning) and based on higher order statistics (BCM learning) have been proposed by investigators.; These theories make different assumptions on the environment and the neural architecture. Two common assumptions on the neural architecture are: The visual pathway has no restriction on the synaptic strength and has only one channel connected with one-type lateral geniculate nucleus (LGN) (on-channel); The visual pathway has only excitatory synaptic strength and has both on- and off-type channels. We showed that these modeling strategies are equivalent.; In addition to the important role of neural architecture in the formation of receptive fields for the cometical cells, the environment that an animal is in also plays an important role. The environment is very noisy. The retinae sometimes receive abnormal inputs. We investigated how a biological neuron could overcome the influence of these inputs.; Beyond the general discussion on the role of the environment, we investigated what kind of receptive fields a cortical cell could develop under natural environment and Principal Component Learning rule (PCA)--a variant of correlation-based learning.; However, a simple cell could develop orientation selectivity before birth or before eye-opening. Unlike natural environment, these environments are noise with no orientation features in them. We argued that the onset of orientation selectivity is caused by the induced structure in the correlation function through the LGN preprocessing. The argument was based on PCA learning.; In order to explain the formation of orientation map, phenomenological approaches and have been used by investigators. We provided an objective function formulism for the synaptic modification rules of a cortical network and identified existing theories. We also provided a connection between the two approaches.