The evolution of learning systems: to Bayes or not to be

Nestor Caticha*, Juan Pablo Neirotti

*Corresponding author for this work

    Research output: Chapter in Book/Published conference outputConference publication

    Abstract

    Bayesian algorithms pose a limit to the performance learning algorithms can achieve. Natural selection should guide the evolution of information processing systems towards those limits. What can we learn from this evolution and what properties do the intermediate stages have? While this question is too general to permit any answer, progress can be made by restricting the class of information processing systems under study. We present analytical and numerical results for the evolution of on-line algorithms for learning from examples for neural network classifiers, which might include or not a hidden layer. The analytical results are obtained by solving a variational problem to determine the learning algorithm that leads to maximum generalization ability. Simulations using evolutionary programming, for programs that implement learning algorithms, confirm and expand the results. The principal result is not just that the evolution is towards a Bayesian limit. Indeed it is essentially reached. In addition we find that evolution is driven by the discovery of useful structures or combinations of variables and operators. In different runs the temporal order of the discovery of such combinations is unique. The main result is that combinations that signal the surprise brought by an example arise always before combinations that serve to gauge the performance of the learning algorithm. This latter structures can be used to implement annealing schedules. The temporal ordering can be understood analytically as well by doing the functional optimization in restricted functional spaces. We also show that there is data suggesting that the appearance of these traits also follows the same temporal ordering in biological systems.

    Original languageEnglish
    Title of host publicationBayesian inference and maximum entropy methods in science and engineering
    EditorsAli Mohammad-Djafari
    PublisherAIP
    Pages203-210
    Number of pages8
    ISBN (Print)978-0-7354-0371-6
    DOIs
    Publication statusPublished - 29 Dec 2006
    EventBayesian inference and maximum entropy methods In science and engineering - Paris, France
    Duration: 8 Jul 200613 Jul 2006

    Publication series

    NameAIP conference proceedings
    PublisherAIP
    Volume872
    ISSN (Print)0094-243X
    ISSN (Electronic)1551-7616

    Conference

    ConferenceBayesian inference and maximum entropy methods In science and engineering
    Country/TerritoryFrance
    City Paris
    Period8/07/0613/07/06

    Fingerprint

    Dive into the research topics of 'The evolution of learning systems: to Bayes or not to be'. Together they form a unique fingerprint.
    • Online learning in discrete hidden Markov models

      Alamino, R. C. & Caticha, N., 29 Dec 2006, Bayesian inference and maximum entropy methods In science and engineering. Mohammad-Djafari, A. (ed.). AIP, p. 187-194 8 p. (AIP conference proceedings; vol. 872).

      Research output: Chapter in Book/Published conference outputConference publication

      Open Access
      File

    Cite this