Minimal Models and Canonical Neural Computations

M. Chirimuuta
April 23. 2013
Minimal Models and Canonical Neural Computations
In a recent paper, Kaplan (2011) takes up the task of extending Craver’s (2007)
mechanistic account of explanation in neuroscience to the new territory of
computational neuroscience (i.e. research using applied mathematics and
computer science to analyze and simulate neural systems). He presents the
model to mechanism mapping (3M) criterion as a condition for a model’s
explanatory adequacy. This mechanistic approach to computational modeling
in neuroscience is intended to replace earlier accounts which posited a level of
computational analysis conceived as autonomous from underlying
mechanistic details (Marr 1982).
In this talk I will discuss work in computational neuroscience that
creates difficulties for Kaplan’s project. Carandini and Heeger (2012) propose
that many neural response properties can be understood in terms of canonical
neural computations. These are “standard computational modules that apply
the same fundamental operations in a variety of contexts.” Importantly, these
computations can have numerous biophysical realisations, and so
straightforward examination of the mechanisms underlying these
computations carries little explanatory weight. Rather than advocate a return
to Marr’s system of independent levels of analysis, I propose that Carandini
and Heeger’s approach should be understood as an instance of minimal
modeling, comparable to that described in other branches of science
(Batterman 2002).