Feature Selection Using Submodular Optimization

Solution Overview

Problem

Existing feature selection techniques fail to guarantee optimal selection from both relevance and redundancy viewpoints, leading to inappropriate feature combinations due to separate optimization of relevance and redundancy, and high calculation costs.

Innovation Solution

An information processing apparatus and method that evaluates feature relevance and redundancy simultaneously using a submodular objective function, optimizing feature selection with a relevance evaluator, redundancy evaluator, and selected feature determiner to achieve appropriate feature selection in polynomial time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all feature combinations are directly evaluated to ensure optimal feature selection, then feature selection accuracy is improved, but calculation cost increases significantly

Engineering Contradiction:

Improvefeature selection accuracyVSAvoidcalculation cost

Core Design Contradiction:

Measurement precisionVSLoss of energy

Solution Approach 1:

The patent segments the feature selection process into two separate optimization stages: relevance optimization and redundancy optimization. This divides the原本 intractable problem of evaluating all feature combinations into two manageable sub-problems that can each be optimized independently, significantly reducing calculation cost while maintaining selection accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary clustering of features based on redundancy relationships before the actual feature selection process. This preliminary action organizes the feature space in advance, allowing the subsequent selection process to work with pre-grouped features rather than individual features, thereby reducing the search space and computational burden.

Inventive Principle:

Principle #10Preliminary action

2Ease of operation

If features are selected one by one based on relevance and redundancy, then selection process is simplified, but the final feature combination may be inappropriate

Engineering Contradiction:

Improveselection process simplicityVSAvoidfeature combination appropriateness

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent performs preliminary clustering of features based on redundancy relationships before the actual feature selection process. This preliminary action organizes the feature space in advance, allowing the subsequent selection process to work with pre-grouped features rather than individual features, thereby reducing the search space and computational burden.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces clustering results as an intermediary structure between the raw feature set and the final selected features. This intermediary representation captures redundancy relationships among features, enabling the selection process to consider both relevance and redundancy simultaneously without evaluating all feature combinations, thus maintaining reliability while simplifying the process.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If relevance and redundancy optimizations are performed separately, then computational complexity is reduced, but truly appropriate feature selection is not guaranteed

Engineering Contradiction:

Improveoptimization complexityVSAvoidfeature selection quality

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary clustering of features based on redundancy relationships before the actual feature selection process. This preliminary action organizes the feature space in advance, allowing the subsequent selection process to work with pre-grouped features rather than individual features, thereby reducing the search space and computational burden.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces clustering results as an intermediary structure between the raw feature set and the final selected features. This intermediary representation captures redundancy relationships among features, enabling the selection process to consider both relevance and redundancy simultaneously without evaluating all feature combinations, thus maintaining reliability while simplifying the process.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS9779207B2Information processing apparatus information processing method, and storage medium

Publication Date: 2017.10.03 NEC CORP

AI summary

An information processing apparatus that selects appropriate features in polynomial time from the viewpoint of both the relevance and the redundancy of features to be selected. This information processing apparatus includes a relevance evaluator that evaluates relevance of each feature included in a set of features, a redundancy evaluator that evaluates redundancy between the features included in the set of features, and a selected feature determiner that determines selected features that optimize a submodular objective function defined using the relevance calculated by the relevance evaluator and the redundancy calculated by the redundancy evaluator.