Christopher D. Green (1998) Lateral Inhibition is a Good Example. Psycoloquy: 9(29) Connectionist Explanation (26)

Volume: 9 (next, prev) Issue: 29 (next, prev) Article: 26 (next prev first) Alternate versions: ASCII Summary
Topic:
Article:
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 9(29): Lateral Inhibition is a Good Example

LATERAL INHIBITION IS A GOOD EXAMPLE
Reply to Bridgeman on Connectionist-Explanation

Christopher D. Green
Department of Psychology
York University
Toronto, Ontario M3J 1P3
Canada
http://www.yorku.ca/faculty/academic/christo/

christo@yorku.ca

Abstract

Bridgeman's (1998) example of a class of networks that are grounded in the known neuroanatomy of the Limulus addresses many of the problems I raised quite nicely. I also discuss the differences between the terms "connectionist," "PDP," and "neural network."

Keywords

artificial intelligence, cognition, computer modelling, connectionism, epistemology, explanation, methodology, neural nets, philosophy of science, theory.
1. Bridgeman's (1998) commentary is very interesting, though there is little for me to reply to. He offers an example of a class of networks grounded in the visual neuroanatomy of the Limulus. As he says, such models have "greater scientific value because they test the properties of specified algorithms and because they instantiate known neuroanatomy" (para. 6). As I have said on several occasions during this treatment, it is not my position that connectionist models MUST be isomorphic to neural structure, though that would seem to be the most obvious aspect of cognitive systems for them to be isomorphic to. In any case, they must be isomorphic of SOMETHING in the cognitive system for us to be able to assess their import. The example Bridgeman raises seems to satisfy that concern.

2. Bridgeman remarks (1998, Abstract) that PDP models are sometimes "misnamed" connectionist. This is not "misnaming," though it courts an ambiguity. "Connectionist" is a broad term that encompasses PDP models, localist network models (such as "semantic networks"), and even Thorndike's old learning theory. Another popular term, "neural network," is to be avoided because it assumes a particular theory of such networks -- viz., that they are simulations of neural structure or activity -- a theory that users of the term are often willing to trade on when convenient, but not always willing to commit to explicitly when push comes to shove. That is to say, it carries a touch of propaganda with it, a bit like old communist states calling themselves "People's Republics." The term "connectionist" rather than "PDP" was used in the target article because it lends itself to the morphology of the language better than PDP. Compare: "connectionists" vs. "PDPers"(?), "connectionistic" vs. "PDPish"(?), "connectionism" vs. "PDPishness"(?)... In any case, my target article was intended mainly as a discussion of PDP models, not of all things that might be described as "connectionist."

REFERENCES

Bridgeman, B. (1998) Models and theories of cognition are algorithms: Commentary on Green on connectionist-explanation. PSYCOLOQUY 9 (22) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.22.connectionist-explanation.19.bridgeman

Green, C.D. (1998) Are Connectionist Models Theories of Cognition? PSYCOLOQUY 9 (4) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.04.connectionist-explanation.1.green


Volume: 9 (next, prev) Issue: 29 (next, prev) Article: 26 (next prev first) Alternate versions: ASCII Summary
Topic:
Article: