Python Programming for Biology: Bioinformatics and Beyond

Figure 24.1.  A simple schematic illustrating the application of machine learning to

the classification problem of text character recognition. A training data set consists of

many known examples for each type of letter, represented as a grid of image pixels with

different intensities. Using the training data, generalised pattern classes representing the

different characters are constructed. This is often achieved by the calculation of decision

boundaries between the different possibilities. The identity of query characters is predicted

by finding the best-matching category pattern. This prediction is often quick because the

match is performed on the generalised class features, rather than comparing to all the

training data.