SP NEURAL
Sharing by
reference versus literal sharing in cell assemblies
Cognition Research Report, 2004
(MS Word):
sharing-by-reference3.doc.
Neural realisation
of the SP theory: cell assemblies revisited
Cognition Research Report, 2004
(MS Word): sbr-functional2.doc.
These papers have not been peer reviewed. Please do not copy or cite without
author's permission.
These two articles are a
radical revision of the one below. The first describes
a serious weaknesses in Hebb's (1949) concept of a cell assembly, namely that
literal sharing of structures between cell assemblies fails to represent
alternative orderings (or spatial configurations) of low-level assemblies. The
paper also shows how this weakness of the cell assembly concept—and several
others identified by Milner (1996)—can be overcome if sharing of structure is
achieved by means of a neuron (or small group of neurons) within each
higher-level assembly that serves as a proxy for or neural reference to the
shared lower-level assembly—so that any one participating neuron belongs in one
assembly and only one assembly.
The second paper describes
functional aspects of the structural proposals in the first paper. Both sets of
ideas are derived from the SP theory of computing and cognition.
Neural
realisation of the SP theory: cell assemblies revisited
Cognition Research Report, 2003 (MS Word):
spn9.doc. Archive:
uk.arxiv.org/abs/cs.AI/0307060.
Previously: Neural mechanisms
for information compression by multiple alignment, unification and search
School of Informatics Report, March 2002, University of Wales Bangor. Archive:
cogprints.ecs.soton.ac.uk/archive/00002824/.
This article, now superceded by the two above, describes how the SP concepts may be
realised in terms of neural mechanisms and neural processing. The proposals in
this article may be seen as an extension and development of Hebb’s[1949]
concept of a ‘cell assembly’. The article describes how the concept of ‘pattern’
in the SP framework may be mapped onto a version of the cell assembly
concept and the way in which neural mechanisms may achieve the effect of ‘multiple
alignment’ in the SP framework. By contrast with the Hebbian concept of
a cell assembly, it is proposed here that any one neuron can belong in one
assembly and only one assembly. A key feature of present proposals, which is
not part of the Hebbian concept, is that any cell assembly may contain ‘references’
or ‘codes’ that serve to identify one or more other cell assemblies. This
mechanism allows information to be stored in a compressed form, it provides a
robust mechanism by which assemblies may be connected to form hierarchies and
other kinds of structure, it means that assemblies can express abstract
concepts, and it provides solutions to some of the other problems associated
with cell assemblies.
Drawing on insights derived from the SP framework, the
article also describes how learning may be achieved with neural mechanisms.
This concept of learning is significantly different from the Hebbian concept
and appears to provide a better account of what we know about human learning.
|