Original Contribution: Training a 3-node neural network is NP-complete
Explore this paper's citation graph
Summary
It is NP-complete to decide whether there exist weights and thresholds for this network so that it produces output consistent with a given set of training examples, and this results suggest that those looking for perfect training algorithms cannot escape inherent computational difficulties just by considering only simple or very regular networks.
- Type
- article
- Published
- 1992-01-05
- Cited by
- 999
- References
- 16
- Access
- Open access
- OpenAlex
- https://openalex.org/W196871588
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:8567973
Keywords
Simple (philosophy), Computer science, Node (physics), Artificial neural network, Set (abstract data type)
References
- Scaling Relationships in Back-propagation Learning
- Improving the performance guarantee for approximate graph coloring
- Computers and Intractability: A Guide to the Theory of NP-Completeness
- Sigmoids Distinguish More Efficiently Than Heavisides
- An O(n0.4)-approximation algorithm for 3-coloring
- Learning in threshold networks
- On the learnability of Boolean formulae
- On the complexity of polyhedral separability
- Neural network design and the complexity of learning
- Cryptographic limitations on learning Boolean formulae and finite automata
- Crytographic limitations on learning Boolean formulae and finite automata
- What Size Net Gives Valid Generalization?
- Learning internal representations by error propagation
- Parallel Networks that Learn to Pronounce English Text
- Neural network design and the complexity of learning
- Neural network design and the complexity of learning
- Sigmoids distinguish better than Heavisides
Cited by
- A Mathematical Solution to a Network Designing Problem
- Estimation of Discriminative Feature Subset Using Community Modularity
- Useful Feature Subsets and Rough Set Reducts
- Cryptographic Hardness Results for Learning Intersections of Halfspaces
- A framework to deal with interference in connectionist systems
- Neural network design and the complexity of learning, by J. Stephen Judd. Cambridge, MA: MIT Press, 1990
- Some Topics in Neural Networks and Control
- Learning and representation: Tensions at the interface
- Presenting and Analyzing the Results of AI Experiments: Data Averaging and Data Snooping
- The neural network loading problem is undecidable
- Induction of Oblique Decision Trees
- Learning Efficiently with Neural Networks: A Theoretical Comparison between Structured and Flat Representations
- Arabic Text Categorization System - Using Ant Colony Optimization-Based Feature Selection
- Some Notes on Computational Learing Theory
- Discovering hierarchical decision rules with evolutive algorithms in supervised learning
- Neural networks and multimedia datasets: estimating the size of neural networks for achieving high classification accuracy
- Learning recursive data is intractable
- Neural network learning of nonstationary processes
- NET PROPHET : McCulloch and developments from his neural net model
- Constructivist neural network models of cognitive development
Related papers
- Training Systems Concept for the Armored Family of Vehicles with Consideration of the Roles of Embedded Training and Stand-Alone Training Devices
- The [0-Simple] Simple Subsemigroups of Nonnegative Matrices
- Fundamental relations in simple and 0-simple semihypergroups of small size
- Effects of Training Input on Training Performance of Airline Service Training - Focused on Mediating Role of Training Process -
- Book Review: IV. Pastorial — Practical Studies: Simple Sermons for Times like These, Simple Sermons on Evangelistic Themes, Simple Talks for Christian Workers, Simple Sermons on Prophetic Themes, Simple Sermons about Jesus Christ, Simple Sermons on Heaven, Hell, and Judgment, Simple Sermons on the Ten Commandments, Simple Sermons for a Sinful Age, Simple Sermons on the Seven Churches of Revelation
- Models of intermodal node representation