Information theory and Visual Plasticity

The relevance of information theory to neural networks has become more apparent in recent years. This theory has become important in analysing and understanding the nature of the neuronal code that is relayed between cortical layers and the nature of the learning goals that guide neuronal learning and synaptic modification. The rapid advance in recent years of single and multiple electrode recording as well as other non-invasive techniques provides a window on neuronal activity and synaptic modification. However, the puzzle is still unsolved; we do not know what are neuronal learning goals, how they are being incorporated and most importantly, what is the nature of the neuronal code and how it is being formed and interpreted by successive layers. When we try to understand the nature of synaptic learning rules, we should be concerned with possible goals that may underly synaptic changes. It is conceivable that knowing what could be a useful goal under different input environments could serve for distinguishing between synaptic plasticity theories. Only after this distinction between goals is indicated, can one continue further and distinguish between learning rules aimed at achieving the same objective, on the basis of their detailed mathematical properties, or computational complexity etc. In this paper, we concentrate on the analysis of different neuronal goals and attempt to provide an update on the current related issues and future directions.