COLT '90 covers the proceedings of the Third Annual Workshop on Computational Learning Theory, sponsored by the ACM SIGACT/SIGART, University of Rochester, Rochester, New York on August 6-8, 1990. The book focuses on the processes, methodologies, principles, and approaches involved in computational learning theory. The selection first elaborates on inductive inference of minimal programs, learning switch configurations, computational complexity of approximating distributions by probabilistic automata, and a learning criterion for stochastic rules. The text then takes a look at inductive identification of pattern languages with restricted substitutions, learning ring-sum-expansions, sample complexity of PAC-learning using random and chosen examples, and some problems of learning with an Oracle. The book examines a mechanical method of successful scientific inquiry, boosting a weak learning algorithm by majority, and learning by distances. Discussions focus on the relation to PAC learnability, majority-vote game, boosting a weak learner by majority vote, and a paradigm of scientific inquiry. The selection is a dependable source of data for researchers interested in the computational learning theory.

Table of Contents

Invited Lecture

Inductive Inference of Minimal Programs

Technical Papers

Identifying μ-Formula Decision Trees with Queries

Learning Switch Configurations

On the Computational Complexity of Approximating Distributions by Probabilistic Automata

A Learning Criterion for Stochastic Rules

On the Complexity of Learning Minimum Time-Bounded Turing Machines

Inductive Inference from Positive Data is Powerful

Inductive Indentification of Pattern Languages with Restricted Substitutions

Pattern Languages Are Not Learnable

On Learning Ring-Sum-Expansions

Learning Functions of k Terms

On the Sample Complexity of Pac-Learning Using Random and Chosen Examples

Finite Learning by a "Team"

Some Problems of Learning with an Oracle

A Mechanical Method of Successful Scientific Inquiry

Boosting a Weak Learning Algorithm by Majority

On the Sample Complexity of Weak Learning

Learning by Distances

The Learnability of Formal Concepts

Polynomial Time Algorithms for Learning Neural Nets

Composite Geometric Concepts and Polynomial Predictability

Learning Integer Lattices

On the Number of Examples and Stages Needed for Learning Decision Trees

Learning DNF Under the Uniform Distribution in Quasi-Polynomial Time

Learning Via Queries with Teams and Anomalies

Learning Via Queries in [+,<]

On the Sample Complexity of Finding Good Search Strategies

Inferring Graphs from Walks

Aggregating Strategies

Short Abstracts

Learning Conjunctions of Horn Clauses

Exact Identification of Circuits Using Fixed Points of Amplification Functions

Efficient Di


No. of pages:
© 1990
Morgan Kaufmann
Print ISBN:
Electronic ISBN: