Abstract
Heterogeneity of functional units for neural computation has been well documented at all the level of hierarchy from individual neurons to entire brain. This heterogeneity has been gained in an adaptive manner, which is far different from the design principle of artificially developed computer. In this review, we first introduce empirical proofs of multi-scale heterogeneity in the neural representation based on our information theoretical analyses in the auditory cortex of rats. Second, our decoding-based analyses investigate how various kinds of neuronal activities are orchestrated in the representation. In light of these results, we discuss about the significance of common characteristics in the sensory cortex, such as the heterogeneity, sparse coding and columnar organization, in terms of an adaptive strategy of neural representation.
