The “Brain-State-in-a-Box” neural model is a gradient descent algorithm

https://doi.org/10.1016/0022-2496(86)90043-XGet rights and content

Abstract

The Brain-State-in-a-Box (BSB) neural model (J. A. Anderson, J. W. Silverstein, S. A. Ritz, & R. S. Jones, 1977, Psychological Review, 84, 413–451) is a pattern categorization device inspired by neurophysiological considerations. This model has additionally been applied to a fairly diverse range of psychological phenomena. In this paper, the BSB model is demonstrated to be a deterministic constrained gradient descent algorithm that minimizes a quadratic cost function. A formal proof that all trajectories of the BSB algorithm in state vector space approach the set of system equilibrium points, under certain specific conditions, is presented. Some conditions regarding the existence of global energy minima are also briefly discussed.

References (12)

  • A. Kawamoto et al.

    A neural network model of multistable perception

    Acta Psychologica

    (1985)
  • J.A. Anderson

    Cognitive and psychological computation with neural models

    IEEE Transactions on Systems, Man, and Cybernetics

    (1983)
  • J.A. Anderson et al.

    Distinctive features, categorical perception, and probability learning: Some applications of a neural model

    Psychological Review

    (1977)
  • R.O. Duda et al.

    Pattern classification and scene analysis

    (1973)
  • R.M. Golden

    A developmental neural model of word perception

  • J.J. Hopfield

    Neurons with graded response have collective properties like those of two-state neurons

There are more references available in the full text version of this article.

Cited by (70)

  • Distributed representations of ambiguous words and their resolution in a connectionist network

    2013, Lexical Ambiguity Resolution: Perspective from Psycholinguistics, Neuropsychology and Artificial Intelligence
  • Information storage and retrieval analysis of hierarchically coupled associative memories

    2012, Information Sciences
    Citation Excerpt :

    On the other hand, the inter-group weight matrix Wcor(a,b) is designed to follow the generalised Hebb rule or Outer Product Method algorithm through the selected patterns extracted from the first-level memories. The energy function studied by Golden [12] can be viewed as being a special case of Eq. (9) when γ = 0 and Nr = 0 (individual network). Golden, in his studies, was able to demonstrate that the network energy decreases over time.

  • Optimal and robust design of brain-state-in-a-box neural associative memories

    2010, Neural Networks
    Citation Excerpt :

    Hopfield (1984) analyzed characteristics of a continuous-time BSB neural network using Lyapunov function approach. Golden (1986) proved that all trajectories of the BSB neural network with a real symmetric weight matrix approach to the set of equilibrium points under a certain condition, which will be used as the global stability condition in the present paper. Li, Michel, and Porod (1989) analyzed a continuous-time BSB neural network, which is referred to as the linear system in a saturated mode (LSSM) model, and investigated the stability of equilibrium points in the LSSM model.

View all citing articles on Scopus

This research was funded in part by Grant BNS-82-14728 from the National Science Foundation, Memory and Cognitive Processes Section to J. A. Anderson.

View full text