Hardness results for neural network approximation problems

Theoretical Computer Science

Volumn 284, Issue 1, 2002, Pages 53-66

  Bartlett, Peter L ^a   Ben David, Shai ^b  

^a AUSTRALIAN NATIONAL UNIVERSITY   (Australia)

^b TECHNION ISRAEL INSTITUTE OF TECHNOLOGY   (Israel)

Author keywords

[No Author keywords available]

Indexed keywords

APPROXIMATION THEORY; COMPUTATIONAL COMPLEXITY; LEARNING SYSTEMS; PROBLEM SOLVING;

LINEAR THRESHOLD NETWORKS;

NEURAL NETWORKS;

EID: 0037031025     PISSN: 03043975     EISSN: None     Source Type: Journal    
DOI: 10.1016/S0304-3975(01)00057-3     Document Type: Conference Paper

References (18)

1. Hardness of approximate optima in lattices, codes, and linear systems
2. On the difficulty of approximately maximizing agreements
3. Training a 3-node neural network is NP-complete
4. On the complexity of training neural networks with continuous activation functions
5. Strong universal consistency of neural network classifiers
6. How to construct random functions
7. Decision theoretic generalizations of the PAC model for neural net and other learning applications
8. Robust trainability of single neurons
9. The computational intractability of training sigmoidal neural networks
10. On the hardness of approximating max-k-cut and its dual
11. Cryptographic limitations on learning Boolean formulae and finite automata
12. Efficient learning of continuous neural networks
13. Efficient agnostic learning of neural networks with bounded fan-in
14. On the complexity of polyhedral separability
15. Optimization, approximation, and complexity classes
16. The hardness of approximation: Gap location
17. On the infeasibility of training neural networks with small squared errors