A Practitioner's Guide to Maximum Causal Entropy Inverse Reinforcement Learning, Starting from Markov Decision Processes

Abstract

This guide is meant to describe both the semantics and mechanics of the Maximum Causal Entropy (MaxCausalEnt) Inverse Reinforcement Learning (IRL) algorithm [4]. Throughout the remainder of this document, we provide a measure of formal definition of the algorithm, starting from the basics, adding some intuition as we go. We intentionally skip a large amount of prior, related, and theoretic work that motivates and contextualizes the algorithm. For further reading on these subjects, see the works referenced throughout. Finally, the gray break out boxes are notes meant to provide broader context, and can be skipped without breaking the ow of the guide.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 2019
Accession Number
AD1090741

Entities

People

  • Eric Heim

Organizations

  • Carnegie Mellon University

Tags

Communities of Interest

  • Ground and Sea Platforms

DTIC Thesaurus Topics

  • Algorithms
  • Anomaly Detection
  • Change Detection
  • Convergence
  • Dynamic Programming
  • Dynamics
  • Frequency
  • Grids
  • Iterations
  • Learning
  • Optimization
  • Probability
  • Probability Distributions
  • Reinforcement Learning
  • Terminals
  • Trajectories
  • Transitions

Fields of Study

  • Computer science

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Mathematical Modeling and Probability Theory.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms