Characteristic simplying method for decision machine

Abstract

A method for feature reduction in a training set for a learning machine such as a Support Vector Machine (SVM). In one embodiment the method includes a step (35) of receiving input training data vectors xi of a training set. The input training data vectors are typically derived from a set of features in a feature space. At step (37) the input data vectors are mapped into a multi-dimensional space. At step (39) a least squares problem, derived from a formulation of the SVM, is solved to determine which features comprising the training vectors are to be deemed significant. At step (41) decision parameters and vectors of the chosen decision machine, e.g. SVM, are determined using the features determined to be significant in step (39).

Description

technical field [0001] The present invention relates to a method of selecting a reduced set of features for a decision machine such as a support vector machine or a principal components analysis system. Background technique [0002] In this specification, reference to prior art is not and should not be considered as an acknowledgment or in any way a suggestion that the prior art forms part of the common general knowledge. [0003] A decision machine is a general learning machine that determines a set of parameters and vectors for classifying unknown data during the training phase. For example, in the case of a support vector machine (SVM), the set of parameters includes a kernel function and a set of support vectors with corresponding multipliers defining a decision hyperplane. The set of support vectors is selected from the population of trained vectors. [0004] Where the decision machine operates according to one of Principal Component Analysis, Kernel Principal Compone...

AI Technical Summary

Problems solved by technology

For example, the word "to" will not be added to the decision of the decision plane, but may appear in many emails

Images