Advertisement

Why Do Machine Learning Based Techniques Fail to Accelerate the Evolution of Neural Networks? Is the Long Bitlength or the Nature of Neural Net Chromosomes to Blame?

  • Hugo de Garis
  • Thayne Batty
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3213)

Abstract

The first author’s primary research interest is in building artificial brains, defined to be interconnected assemblages of 10,000s of electronically evolved neural network circuit modules, whose neural signaling is run in real time in an ordinary PC, to control autonomous robots, etc. An alternative to the electronic evolution of the NN modules is to use software based machine learning techniques in a PC, for example Michalski’s LEM algorithm [Michalski 2000]. LEM can be very successful at accelerating the evolutionary optimization of multivariable mathematical functions, but appears to fail to accelerate the evolution of neural networks [Aleti & de Garis 2004]. This paper reports on experiments comparing the evolution times of mathematical optimization problems with those of neural network evolution problems, where the bit string chromosomes of both sets of problems are the same, to see if there is something special about neural network chromosomes that make them unsuitable to have their evolution accelerated.

Keywords

Chromosome Length Mathematical Optimization Traditional Genetic Algorithm Rule List Neural Network Evolution 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [Aleti & de Garis, 2004]
    Aleti, S.H., de Garis, H.: Evolutionary Algorithms Based on Machine Learning Accelerate Mathematical Function Optimization but not Neural Net Evolution, http://www.cs.usu.edu/~degaris/papers/CEC-2004Sree.pdf
  2. [de Garis & Batty, 2004]
    de Garis, H., Batty, T.: MULTI-MOD: A PC Based Software System for Controlling an Artificial Brain Containing 10,000 Evolved Neural Net Modules. In: Congress on Evolutionary Computation, CEC 2004 (2004), http://www.cs.usu.edu/~degaris/papers/2004-CEC-Multi-Mod.pdf
  3. [de Garis, 2002]
    de Garis, H., Korkin, M.: THE CAM-BRAIN MACHINE (CBM) An FPGA Based Hardware Tool which Evolves a 1000 Neuron Net Circuit Module in Seconds and Updates a 75 Million Neuron Artificial Brain for Real Time Robot Control. In: de Garis, H. (ed.) Neurocomputing journal, February, 2002. Special issue on Evolutionary Neural Systems, vol. 42(1-4), Elsevier, Amsterdam (2002), http://www.cs.usu.edu/~degaris/papers/evolnrlsys.pdf Google Scholar
  4. [Michalski, 2000]
    Michalski, R.S.: Learnable Evolution Model: Evolutionary Processes Guided by Machine Learning. Machine Learning 38, 9–40 (2000)zbMATHCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Hugo de Garis
    • 1
  • Thayne Batty
    • 1
  1. 1.Brain Builder Group, Computer Science DepartmentUtah State UniversityLoganUSA

Personalised recommendations