Fitness Distance Correlation of Neural Network Error Surfaces: A Scalable, Continuous Optimization Problem

Abstract

This paper investigates neural network training as a potential source of problems for benchmarking continuous, heuristic optimization algorithms. Through the use of a student-teacher learning paradigm, the error surfaces of several neural networks are examined using so-called fitness distance correlation, which has previously been applied to discrete, combinatorial optimization problems. The results suggest that the neural network training tasks offer a number of desirable properties for algorithm benchmarking, including the ability to scale-up to provide challenging problems in high-dimensional spaces.

Cite

Text

Gallagher. "Fitness Distance Correlation of Neural Network Error Surfaces: A Scalable, Continuous Optimization Problem." European Conference on Machine Learning, 2001. doi:10.1007/3-540-44795-4_14

Markdown

[Gallagher. "Fitness Distance Correlation of Neural Network Error Surfaces: A Scalable, Continuous Optimization Problem." European Conference on Machine Learning, 2001.](https://mlanthology.org/ecmlpkdd/2001/gallagher2001ecml-fitness/) doi:10.1007/3-540-44795-4_14

BibTeX

@inproceedings{gallagher2001ecml-fitness,
  title     = {{Fitness Distance Correlation of Neural Network Error Surfaces: A Scalable, Continuous Optimization Problem}},
  author    = {Gallagher, Marcus},
  booktitle = {European Conference on Machine Learning},
  year      = {2001},
  pages     = {157-166},
  doi       = {10.1007/3-540-44795-4_14},
  url       = {https://mlanthology.org/ecmlpkdd/2001/gallagher2001ecml-fitness/}
}