|
Information theory provides a natural mathematical framework to quantify how much information is contained in the firing patterns. This methodology has been extensively used to study neural coding in different brain areas. Quantitative estimation of mutual information requires large amount of data and different techniques have been developed to overcome the limited sampling problem.
Involved in sensorimotor and even cognitive coordination, the cerebellum receives inputs from diverse brain areas. These inputs are recoded by the granule cells, the most numerous cells in the brain, and together with the inhibitory Golgi cells, form the granular layer.
We are investigating information transfer in the cerebellar granule cell layer with special attention to the mossy fibre-granule cell relay. This system is characterized by a small number of synaptic inputs (4-7) allowing one to dissect the noise sources during neurotransmission and to carry out an extensive analysis of the input stimuli.
Computing stimulus specific information (SSI), we were able to identify the most informative set of stimuli, and to understand in which conditions cell coding capability is optimized. This kind of information-theoretic analysis constitutes an increasingly important tool to characterize non-linearities in neural processing.
Although simple, we also suggest that the granular layer contains different types of signal processing that leads to a robust and efficient representation for coordination. We have shown that the granular layer has the possibility, with a very simple network, to perform a coding 1) that is efficient: cells are active only when there is something to code, 2) that facilitates learning: independent features are coded by different cells, 3) that performs optimal denoising, 4) that is robust to decreases in the number of coding cells.
|
![[SONY]](../../../inline/badgeSony.jpg){kind=small}
