CODEWITHSUNDEEP
algorithmmathmachine-learningpattern-recognition
Simplex
Simplex

Reputation: 1733

Pattern recognition: one class only

I've got a question: how can I do pattern recognition, if I have only one class? In other words, I want get answer from algorithm is it my pattern, or not, i. e. 1 or 0.

At first sight, it is a simple two classes classification, but in this case arise problem: what is class, that describe 0 class (how we can describe complement to universum?) If set A - set, that contains objects with some property, then not A (U\A, where U is universum) - set of objects, without this property.

I can get examples this task in particularly: Face recognition (it is face of person, or not) Recognition some thing on the picture. In this case in one piece training set I must specify pictures with person face (or some thing), but what must contain another piece training set? I can't specify all pictures, that will not contain this thing.

Upvotes: 1

Views: 183

Answers (1)

amit
amit

Reputation: 178441

This is simple classification problem, the question is "Is it an item from the set?" and the question has a boolean answer: true or false.

In the example of face recognition - you are actually looking for face detection. This can be done, for example using viola-jones object detection framework.

The idea is to provide items that are in the set (faces), and some randomly chosen items that are not in the set. and classify them with the boolean classifier.

Upvotes: 4

Related Questions