A minimum description length principle for perception
UNSPECIFIED (2005) A minimum description length principle for perception. In: Workshop on Advances in Minimum Description Length - Theory and Applications, Whistler, CANADA, DEC, 2001. Published in: Advances in Minimum Description Length Theory and Applications pp. 385-409.Full text not available from this repository.
Perception involves inferring the structure of the environment fromsensory data. As in any process of inductive inference, there are infinitely many hypotheses about environmental structure that are compatible with sensory data. Can the perceptual system use the minimum description length (MDL) principle, which prefers hypotheses that provide a short explanation of the data, to choose between these competing explanations? This viewpoint has a long history in psychology, which can be traced back to Ernst Mach and to Gestalt psychology. This chapter considers how the MDL approach relates to apparently rival principles of perception, what types of empirical data the approach can and cannot explain, and how an MDL approach to perception might be augmented to provide an empirically adequate framework for understanding perceptual inference.
|Item Type:||Conference Item (UNSPECIFIED)|
|Subjects:||Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software
Q Science > QA Mathematics
|Series Name:||NEURAL INFORMATION PROCESSING SERIES|
|Journal or Publication Title:||Advances in Minimum Description Length Theory and Applications|
|Publisher:||M I T PRESS|
|Editor:||Grunwald, PD and Myung, IJ and Pitt, MA|
|Number of Pages:||25|
|Page Range:||pp. 385-409|
|Title of Event:||Workshop on Advances in Minimum Description Length - Theory and Applications|
|Location of Event:||Whistler, CANADA|
|Date(s) of Event:||DEC, 2001|
Actions (login required)