Philosophical aspects in pattern recognition research

Abstract

Pattern recognition is the discipline which studies theories and methods to build machines that are able to discover regularities in noisy data. As many of its characterizations suggest, the field is usually considered as a purely engineering discipline. But, in a broad sense, pattern recognition has in fact an interdisciplinary nature. Surprisingly, the attention towards philosophical issues and foundations, which marked the early days of the field, in time has fallen into oblivion. In this thesis, we aim to recover this original interdisciplinary attitude by discussing the philosophical underpinnings of today’s pattern recognition research. First we will approach the question of the very nature of the field thanks to the recent developments of both the philosophy of technology and the philosophy of science. This will bring us to consider pattern recognition problems from a slightly different perspective, that is by focusing their relations to some cognitive phenomena (and, in particular, to categorization). Finally, we will undertake a critical analysis of the main research tendencies by making use of Kuhn’s notion of a paradigm, one of the cornerstones of the twentieth-century philosophy of science.