CiteSeerX — Training A 3-Node Neural Network Is NP-Complete

Training A 3-Node Neural Network Is NP-Complete (1992) [175 citations — 2 self]

Popular Tags

No tags have been applied to this document.

BibTeX | Add To MetaCart

@MISC{Blum92traininga,
    author = {Avrim L. Blum and Ronald L. Rivest},
    title = {Training A 3-Node Neural Network Is NP-Complete},
    year = {1992}
}

Bookmarks

OpenURL

Abstract:

: We consider a 2-layer, 3-node, n-input neural network whose nodes compute linear threshold functions of their inputs. We show that 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. We extend the result to other simple networks. We also present a network for which training is hard but where switching to a more powerful representation makes training easier. These results suggest that those looking for perfect training algorithms cannot escape inherent computational difficulties just by considering only simple or very regular networks. They also suggest the importance, given a training problem, of finding an appropriate network and input encoding for that problem. It is left as an open problem to extend our result to nodes with non-linear functions such as sigmoids. Keywords: Neural networks, computational complexity, NP-completeness, intractability, learning, training, mu...

Citations

7271	Computers and Intractability - A Guide to the Theory of NP-Completeness – Garey, Johnson - 1979
2044	Learning internal representations by error propagation – Rumelhart, G, et al. - 1986
383	Parallel networks that learn to pronounce English text – Sejnowski, Rosenberg - 1987
281	What size net gives valid generalization – Baum, Haussler - 1989
242	Cryptographic limitations on learning boolean formulae and finite automata – Kearns, Valiant - 1994
140	On the learnability of Boolean formulae – Kearns, Li, et al. - 1987
82	Neural Network Design and the Complexity of Learning", MIT-Press – Judd - 1990
57	Improving the Performance Guarantee for Approximate Graph Coloring – Wigderson - 1983
34	On the complexity of polyhedral separability – Megiddo - 1996
34	Scaling Relationships in Back-propagation Learning: Dependence on Predicate Order – Tesauro, Janssens - 1988
33	Training a 3-node neural net is NP-Complete – Blum, Rivest - 1989
18	Generalizing the PAC model for neural net and other learning applications – Haussler - 1989
13	An ~ O(n 0:4 )-approximation algorithm for 3-coloring (and improved approximation algorithms for k-coloring – Blum - 1989
13	Learning in threshold networks – Raghavan - 1988
2	On the computational complexity of training simple neural networks. Master 's thesis – Blum - 1989
2	Sigmoids distinguish better than Heavisides – Sontag - 1989
1	Predicting symmetric differences of two halfspaces reduces to predicting halfspaces. Unpublished manuscript – Valiant, Warmuth - 1989