Paul E. Utgoff

JMLR Journal 2004 Journal Article

Randomized Variable Elimination

• David J. Stracuzzi
• Paul E. Utgoff

Variable selection, the process of identifying input variables that are relevant to a particular learning problem, has received much attention in the learning community. Methods that employ a learning algorithm as a part of the selection process (wrappers) have been shown to outperform methods that select variables independently from the learning algorithm (filters), but only at great computational expense. We present a randomized wrapper algorithm whose computational requirements are within a constant factor of simply learning in the presence of all input variables, provided that the number of relevant variables is small and known in advance. We then show how to remove the latter assumption, and demonstrate performance on several problems. [abs] [ pdf ] [ code ]

ICML Conference 2002 Conference Paper

Randomized Variable Elimination

• David J. Stracuzzi
• Paul E. Utgoff

ICML Conference 1999 Conference Paper

Approximation Via Value Unification

• Paul E. Utgoff
• David J. Stracuzzi

ICML Conference 1998 Conference Paper

Classification Using Phi-Machines and Constructive Function Approximation

• Doina Precup
• Paul E. Utgoff

ICML Conference 1994 Conference Paper

An Improved Algorithm for Incremental Induction of Decision Trees

• Paul E. Utgoff

AAAI Conference 1991 Conference Paper

Two Kinds of Training Information For Evaluation Function Learning

• Paul E. Utgoff

This paper identifies two fundamentally different kinds of training information for learning search control in terms of an evaluation function. Each kind of training information suggests its own set of methods for learning an evaluation function. The paper shows that one can integrate the methods and learn simultaneously from both kinds of information.

AAAI Conference 1990 Conference Paper

Explaining Temporal Differences to Create Useful Concepts for Evaluating States

• Richard C. Yee
• Paul E. Utgoff

We describe a technique for improving problemsolving performance by creating concepts that allow problem states to be evaluated through an efficient recognition process. A temporakdiflerence (TD) method is used to bootstrap a collection of useful concepts by backing up evaluations from recognized states to their predecessors. This procedure is combined with explanation- based generalization (EBG) and goal regression to use knowledge of the problem domain to help generalize the new concept definitions. This maintains the efficiency of using the concepts and accelerates the learning process in comparison to knowledge-free approaches. Also, because the learned definitions may describe negative conditions, it becomes possible to use EBG to explain why some instance is not an example of a concept. The learning technique has been elaborated for minimax gameplaying and tested on a Tic-Tat-Toe system, T2. Given only concepts defining the end-game states and constrained to a two-ply search bound, experiments show that T2 learns concepts for achieving near-perfect play. T2’ s total searching time, including concept recognition, is within acceptable performance limits while perfect play without the concepts requires searches taking well over 100 times longer than T2’ s.

AAAI Conference 1988 Conference Paper

Perceptron Trees: A Case Study in Hybrid Concept Representations

• Paul E. Utgoff

AAAI Conference 1982 Conference Paper

Acquisition of Appropriate Bias for Inductive Concept Learning

• Paul E. Utgoff

Current approaches to inductive concept learning suffer from a fundamental difficulty; if a fixed language is chosen in which to represent concepts, then in cases where that language is inappropriate, the new concept may be impossible to describe (and therefore to learn). We suggest a framework for automatically extending the language in which concepts are to be expressed. This framework includes multiple sources of knowledge for recommending plausible language extensions.

IJCAI Conference 1981 Conference Paper

A Plot Understanding System on Reference to Both Image and Language

• 77 Metaphor Interpretation as Selective Inferencing 85 -LEARNING 1 - A Computer Model of Child Language Acquisition Mallory Selfridge 92 A Theory of Language Acquisition Based on 97 General Learning Principles 104 Concept Learning by Experiment 104 Analogy-Based Acquisition of Utterances Relating to Temporal Aspects 106 -LEARNING 2- Inductive Learning of Pronunciation Rules by Hypothesis Testing
• Correction S. Oakey
• R. C. Cawthorn 109 Failure-Driven Reminding for Incremental Learning 115 BACON. 5: The Discovery of Conservation Laws Pat Langley
• Gary L. Bradshaw
• 121 -LEARNING 3- Learning Problem-Solving Heuristics Through Practice Tom M. Mitchell
• Paul E. Utgoff