Transductive feature selection with maximum-relevancy and minimum-redundancy criteria

Patent metadata

Field	Value
Publication number	US-9483739-B2
Application number	US-201314030708-A
Country	US
Kind code	B2
Filing date	Sep 18, 2013
Priority date	Jan 21, 2013
Publication date	Nov 1, 2016
Grant date	Nov 1, 2016

Abstract

Various embodiments select features from a feature space. In one embodiment, a set of training samples and a set of test samples are received. The set of training samples includes a set of features and a class value. The set of test samples includes the set of features absent the class value. A relevancy with respect to the class value is determined for each of a plurality of unselected features based on the set of training samples. A redundancy with respect to one or more of the set of features is determined for each of the plurality of unselected features in the first set of features based on the set of training samples and the set of test samples. A set of features is selected from the plurality of unselected features based on the relevancy and the redundancy determined for each of the plurality of unselected features.

First claim

What is claimed is:

1. An information processing system for selecting features from a feature space, the information processing system comprising:
a memory;
a processor communicatively coupled to the memory; and
a feature selection module communicatively coupled to the memory and the processor, wherein the feature selection module is configured to perform a method comprising:
obtaining, by a processor, a set of training samples and a set of test samples, wherein the set of training samples comprises a first set of features and a class value, and wherein the set of test samples comprises a second set of features, where the second set of features is the first set of features absent the class value;
determining, for each of a plurality of unselected features in a plurality of features comprising the first and second set of features, a relevancy with respect to the class value based on the set of training samples;
determining, for each of the plurality of unselected features, a redundancy with respect to the plurality of features based on both the set of training samples and the set of test samples;
selecting a set of features from the plurality of unselected features based on the relevancy and the redundancy determined for each of the plurality of unselected features, wherein the selecting is performed based on;

max

x
j

∈

X
-

S

m
-
1

⁢

[

I
⁡

(

x
j
training

;

c
training

)

-

1

m
-
1

⁢

∑

x
i

∈

S

m
-
1

⁢

I
⁡

(

x
j

training
+
test

;

c

training
+
test

)

]

,

 where x j is a jth feature that is sample independent, x training is a jth feature based on the set of training samples, x j training+test is a jth based on the set of training samples and the set of test samples, i is an integer, X is a set of all features, S m-1 is a set of m−1 features, c is the class value, and I is mutual information; and
programming a processor to perform at least one of a set of classification operations and a set of regression operations based on the set of features that have been selected.

2. The information processing system of claim 1 , wherein each of the set of features that has been selected has a maximum relevancy among each of the plurality of unselected features with respect to the class value based on the set of training samples, and has a minimum redundancy among each of the plurality of unselected features with respect to the set of features based on the set of training samples and the set of test samples.

3. The information processing system of claim 1 , wherein the relevancy is determined based on mutual information between a given unselected feature in the plurality of unselected features and the class value based on the set of training samples.

4. The information processing system of claim 3 , wherein the mutual information is determined based on comprising:
determining that the class value is within a given threshold;
rounding the class value; and
multiplying the class value by a scalar, wherein an entropy of the class value after being multiplied by the scalar is within a given threshold of an original entropy of the class value.

5. The information processing system of claim 1 , wherein the redundancy is determined based on mutual information between a given unselected feature in the plurality of unselected features and each feature in the plurality of features.

6. The information processing system of claim 5 , wherein the method further comprises:
storing a set of counts for each of a set of values used to determine the mutual information between the given unselected feature and the plurality of features; and
determining the mutual information between the given unselected feature and the plurality of features based on the set of counts that has been stored.

7. A non-transitory computer program product for selecting features from a feature space, the computer program product comprising:
a storage medium readable by a processing circuit and storing instructions for execution by the processing circuit for performing a method comprising:
obtaining, by a processor, a set of training samples and a set of test samples, wherein the set of training samples comprises a first set of features and a class value, and wherein the set of test samples comprises a second set of features, where the second set of features is the first set of features absent the class value;
determining, for each of a plurality of unselected features in a plurality of features comprising the first and second set of features, a relevancy with respect to the class value based on the set of training samples;
determining, for each of the plurality of unselected features, a redundancy with respect to the plur

Assignees

• IBM

Inventors

• Haws David
• He Dan
• Parida Laxmi P
• Rish Irina

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/2115

  by evaluating different subsets according to an optimisation criterion, e.g. class separability, forward selection or backward elimination · CPC title

• G06K9/6256

  Physics · mapped topic

• G06N5/025

  Extracting rules from data · CPC title