ABSTRACT
The two central concepts of information theory are those of entropy and information. If think of the early parts of a perceptual system such as vision as a system for transmitting information about the environment on to higher centres, it seems reasonable that the more information which is transmitted, the more effective the system will be. Of course, some visual systems are optimized to extract information about very specific stimuli early on: an example would be the ‘bug detectors’ found in the frog. Principal-components analysis is widely used for dimension reduction in data analysis and pre-processing, and is used under a variety of names such as the Karhunen Loève transform, factor analysis, or the Hotelling transform in image processing. Its primary use is to provide a reduction in the number of parameters used to represent a quantity, while minimizing the error introduced by doing so.
