Posted in Presentation Slides: Tuesday, June 7, 2011
The brain can be viewed as a complicated computational machine. It consists of several billion neurons, each of which functions as a nonlinear, dynamical analog-to-digital filter. These neurons are highly interconnected, and they form distinct functional circuits and subsystems that are organized both serially and in parallel. Furthermore, these complex networks are modulated by feedback circuitry and by secondary non-neuronal systems. Although information is commonly assumed to be transmitted between neurons in terms of a short-term spike rate code, neural computation in the brain is widely distributed over space and time, and the principles of information coding are poorly understood.
Since Shannon's early work in the 1950's the brain has commonly been viewed as an information processing device whose function should in principle be describable in term of information theory (see Borst & Theunissen, 1999). It is clear, however, that the brain is functionally quite different from the communication systems that motivated Shannon's work. Thus far direct application of classical information theory to neuroscience data has not proved to be particularly useful. This is likely because, until quite recently, the quantity and quality of data that could be acquired in neuroscience experiments has been limited. Neurophysiological recordings could be made from only a handful of neurons, and noninvasive approaches such as EEG produced poor data that were difficult to analyze and interepret. Recent technical advances on many fronts have begun to change this picture. Neural data are now much easier to obtain due to advances such as calcium imaging and large-scale parallel neurophysiological recording. And the rapid development of functional MRI has provided a sensitive noninvasive method for obtaining data from the human brain.
These vast new sources of neuroscience data provide many opportunities for application of information theory. However, understanding the brain as an information processing device will require fundamental advances in information theory in order to account for the complex topology, extensive feedback and unknown principles of communication and coding in the brain.