Automated Feature Selection for Experience-Based Adaptive Re-planning

Abstract

This project proposes innovative methodologies of automatic feature selection for Experience-Based Adaptive Replanning (EBAR). EBAR is an extension of the Distributed Episodic Exploratory Planning (DEEP) in-house project conducted at the Air Force Research Lab. This project pursues two complementary research objectives: (i) developing efficient feature selection algorithms for case based planning (CBP), and (ii) evaluating and demonstrating the effectiveness of feature selection for CBP. In this project, we have successfully developed two major types of feature selection methods for CBP, wrapper and filter, which differ mainly in how they evaluate the quality of a feature subset. We have also developed an efficient result validation procedure and demonstrated the efficiency and efficacy of the proposed feature selection methods based on the StarCraft domain.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2013
Accession Number
ADA582125

Entities

People

  • Lei Yu

Organizations

  • Binghamton University

Tags

Communities of Interest

  • Autonomy
  • C4I
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Algorithms
  • Automatic
  • Buildings And Structures
  • Feature Selection
  • Information Operations
  • Military Research
  • Military Science
  • Random Variables
  • Sparse Matrix
  • Statistics
  • Supervised Machine Learning
  • Test Sets
  • Training
  • United States
  • Validation

Fields of Study

  • Computer science

Readers

  • Artificial Intelligence
  • Distributed Systems and Data Platform Development
  • Psychometric Testing or Psychological Assessment.