Many of the problems, what one likes to term in Computer Science machine learning, may be formulated as follows: Assume samples ωi, i=0,1,... drawn after another from some set Ω, which is, say, an Euclidean space or some subset of it. Members ω∈ω may directly characterize results of actual measurements, observations, objects or symptoms. However, it is more appropriate to think of an Ω that is the collection of features we have derived from such data by means of some appropriate many-to-one mapping. (Many of the interesting and crucial techniques of pattern recognition are concerned just with such feature extraction procedures. However, in what follows, the actual interpretation of Ω does not make much matter.)
KeywordsMachine Learn Learning Problem Decision Function Basic Notion Target Quantity
Unable to display preview. Download preview PDF.