Local Structure Learning and Knowledge Augmentation Learning

Abstract

Local structure learning is to discover the local structure for a target variable. A key concept for local structure is the Markov Blanket (MB). The MB of a target variable consists of those variables, which jointly contain all the information needed to predict the behaviors of the target variable. MB learning is to discover the MB of the target node. The research objectives are: 1) development of robust and efficient methods for MB learning; 2) application of the MB learning to various machine learning tasks, including structured feature selection, classification, and causal structure learning, and 3) incorporation of the MB learning methods into D3M environment and evaluation of their performance on D3M datasets.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 23, 2021
Accession Number
AD1153459

Entities

People

  • Qiang Ji

Organizations

  • Rensselaer Polytechnic Institute

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Air Force
  • Air Force Research Laboratories
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Bayesian Networks
  • Computational Science
  • Computer Vision
  • Computers
  • Data Mining
  • Data Science
  • Dimensionality Reduction
  • Feature Selection
  • Governments
  • Information Processing
  • Information Science
  • Information Systems
  • Machine Learning
  • Monte Carlo Method
  • Neural Networks
  • Probability
  • Probability Distributions
  • Random Variables

Fields of Study

  • Computer science

Readers

  • Atmospheric Science/Meteorology
  • Computational Modeling and Simulation
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks