
|

|
 |
The central theme of my 1991 doctoral thesis under the guidance
of Elie Bienenstock was the relationship
between neural code and mental representation.
If we make the assumption
that all mental "entities" (sensation, perception, concept,
word, external object, action, etc.) are represented in the
nervous system as states of neuronal activities, a fundamental
problem of cognitive neuroscience is to elucidate the
structure and properties of such representational states.
|
I conducted three different, yet interrelated studies
advocating Christoph von der Malsburg's theory of temporal
correlations as the basis of the neural code: a handwritten
character classifier (see
), a model
of cortical self-organization (see
),
and a review of the limits of statistical learning in neural
networks (see ).
I developed these various theoretical and practical models and
carried out numerical simulations with the purpose of
illustrating the relevance of the alternative neural code
originally proposed by von der Malsburg.
This question has often been debated since the beginnings of
modern neuroscience but it is generally accepted that the
average firing rate of neurons constitutes an important part
of the neural code. In short, the classical view holds that
mental entities are coded by cell assemblies (Hebb, 1949),
which are spatial patterns of average activity.
Following Christoph von der Malsburg's "Correlation theory
of brain function" (1981) and the work of my thesis advisor,
Elie Bienenstock (see, e.g., von der Malsburg and Bienenstock,
1986), I defended another format of representation that
involves higher-order moments or temporal correlations
among neuronal activities. Here, mental representations are
not exclusively based on individual mean activity rates
<xi>, which are events
of order 1, but more generally on order-N events
<xi1 xi2 ... xiN>
and, in particular, correlations between two neurons,
<xi xj>.
Naturally, the traditional order-1 code stems from classical
observations in the primary sensory areas (e.g., visual cortex),
in which cells seem to possess selective response properties.
From these experiments, it was inferred that one
such neuron, or a small cluster of neurons, could individually
and independently represent one specific type of stimulus
(e.g., the orientation of an edge).
However, to obtain the global representation of an object
these local features must eventually be linked and integrated.
The problem is that this integration is unlikely to be carried
out by highly specialized cells at the top of a hierarchical
processing chain (the conjectural "grand-mother" cells that fire
only when you see your grand-mother). Equally unlikely would
be for the assembly of feature-coding activity rates to be
maintained in a distributed state because of the impossiblity
to overlap two such states without losing relational
information (the so-called "binding problem"). If two cells
coding for "red" and "circle" are active and two other cells
coding for "green" and "triangle"
also become active, then this global state of activation is
undistinguishable from the alternative combination "red triangle"
and "green circle" (von der Malsburg, 1987).
This is why we advocated the idea that feature integration
requires higher-order codes to be able to represent
relationships between elementary components that are initially
uncorrelated (in the above example the spike trains of "red"
and "circle" would be synchronous and out of phase with
"green" and "triangle"). These correlation events bring to
the representation format a structure that is fundamentally
missing from the mere feature lists of Hebbian cell assemblies.
To use a chemical metaphor, we could say that feature lists
are to molecular formulas (e.g., C3H8O)
what correlations are to line-bond diagrams (e.g., 1-propanol
vs. 2-propanol).
The first part does not offer a new method or algorithm but
rather aims at bringing to light general problems and
limitations encountered by statistical learning processes,
especially of the generalist or "nonparametric" kind. The
main goal of this study is to stress the crucial importance
of identifying the right format of representation and
giving it priority over other concerns about the "adaptability"
or power of generalization of a learning system.
We put into practice this recommendation in the second part
by designing a handwritten character classification method
based on order-2 correlations. Images are represented by 2-D
deformable lattices instead of unstructured lists of pixels,
while the "distance" between two input images is defined as the
cost-functional of a graph-matching process. The success rates
achieved by this criteria are superior to feed-forward
neural classifiers, implicitly based on Hamming or Euclidean
metrics. This study illustrates both the importance of a
problem-specific format of representation and the particular
appropriateness of higher-order codes in this matter.
The third part is more speculative and attempts to answer
Fodor and Pylyshyn's (1988) influential criticism about the lack
of structured representations in neural networks. We show that
the compositionality of language and cognition can
actually arise from the simultaneous self-organization of
connectivity and activity in an initially random cortical
network.
|
References
von der Malsburg (1981) Correlation theory of brain function.
MPI report.
von der Malsburg & Bienenstock (1986)
In Disordered syst. and biological org.,
Springer: 247-272.
|
Publications
Bienenstock, E. & Doursat, R. (1990)
Spatio-temporal coding and the compositionality of cognition.
In Proceedings of the Workshop on Temporal Correlations and Temporal Coding in the Brain,
April 25-27, 1990, Paris, France; R. Lestienne, ed.: pp. 42-47.
Doursat, R. (1991)
Contribution à l'étude des représentations dans le
système nerveux et dans les réseaux de neurones formels.
PhD Thesis, Université Pierre et Marie Curie (Paris 6).
|
|
|
The bias/variance dilemma in formal neural networks
|
 |
The first part, in collaboration with
Stuart Geman, does not offer a new method
or algorithm but rather aims at bringing to light general
problems and limitations encountered by statistical learning
processes, especially of the generalist or "nonparametric" kind.
The main goal of this study is to stress the crucial importance
of identifying the right format of representation and
giving it priority over other concerns about the "adaptability"
or power of generalization of a learning system.
It became an oft-cited paper published in Neural Computation
in 1992.
|
In this work, we addressed the issue of representation within the
framework of statistical estimation theory. During the
renewal of interest for connectionist models in the 1980's,
the great majority of neural network methods focused
on classification or estimation problems, especially regression.
These generalist systems are preoccupied with interpolating
sample data through gradual statistical approximation, without
prior or built-in knowledge of the problem at hand. This is
also called nonparametric inference.
Nonparametric estimation has been successful in numerous
application domains and performs in some cases better than
logical inference (expert systems, AI). However, it must also
be asked whether this statistical approach is equally relevant
to the cognitive and neurobiological domains. Does it provide
an answer to the issue of neural representation? Can it solve
complex cognitive problems such as invariant perception or
language?
In short, when dealing with the nervous system and attempting
to unravel the deep mechanisms of cognition, one may question
whether the learning paradigm that drives the core of
these methods is effectively as crucial as it is often claimed.
We show that the efficacy of any statistical estimator is
bounded by universal limits. The mean quadratic error produced
by an estimator (averaged over different learning data sets)
is made of two terms, the bias and variance, which
clearly point to the cause of these limitations. The bias term
represents the average discrepancy between the system's predictions
and the expected response, while the variance represents the
fluctuations of these predictions with the various sample sets.
Our conclusion is that statistical estimation can yield good
results in "simple" problems that do not require the system to
extrapolate to unknown examples in a nontrivial way. Otherwise,
if one expects a high power of generalization after the first
interpolation phase, we claim that it becomes necessary to
design into the system an adequate format of representation
for the data and that the quality of this representation must
have priority over tuning the learning parameters or selecting
the examples. This is especially critical in "complex" problems
of a cognitive nature.
This view is now widely accepted, yet it is still hoped in
many cases that learning methods can actually discover by
themselves the relevant data representation. In particular,
these methods should be able to extract from the examples
hidden component features that are characteristic of their class.
This hope, again, has to be measured against the complexity
of the problem.
|
Publications
Cited over 2400 times (as per Google Scholar):
Doursat, R. (1991)
Contribution à l'étude des représentations dans le
système nerveux et dans les réseaux de neurones formels.
PhD Thesis, Université Pierre et Marie Curie (Paris 6).
|
|
|
Elastic matching for handwritten character recognition
|
 |
We put into practice the first part's recommendation in the
second part by designing a handwritten character classification
method based on order-2 correlations. Images are represented by
2-D deformable lattices instead of unstructured lists of pixels,
while the "distance" between two input images is defined as the
cost-functional of a graph-matching process. The success rates
achieved by this criteria are superior to feed-forward
neural classifiers, implicitly based on Hamming or Euclidean
metrics.
|
In this part we described a concrete implementation of a
shape recognition model inspired by von der Malsburg (1981).
This author offers an original representation format in the
nervous system based on an order-2 neural coding. In this
coding, relationships between objects in a visual scene are
represented by temporal correlations between neuronal activities.
The present study illustrates both the importance of a
problem-specific format of representation and the particular
appropriateness of higher-order codes in this matter.
Von der Malsburg's theory was prompted by the realization
that the average rate-based conceptual format inevitably led
to great difficulties or the impossibility to handle complex
cognitive problems, including visual perception. In short,
the classical Hebbian "cell assembly" format is lacking the
structure necessary to code for relational information (see
).
Therefore, in this framework, the only way to represent
relationships and avoid the "superposition catastrophe" (see
the "red circle/green triangle" example of the introduction)
is to dedicate new cells for each new combination of features.
This solution is obviously not realistic and it is much more
natural to code relations with temporal correlations
among cellular activities rather than with new cells. Moreover,
through a Hebbian-like positive feedback loop on the millisecond
timescale, these correlations reinforce the synaptic
connections that support them. Therefore, instead of
correlations, a relational structure can be equivalently
represented by dynamical links.
|
|
We developed here a shape-recognition algorithm directly
motivated by these principles but also simplified and made more
computationally efficient for practical purposes. The core of
the model is a template matching operation between two labeled
graphs that are relational representations of two shapes.
This graph-matching procedure is construed as a functional
optimization that searches the best (oriented) mapping from
the nodes of the first object to the nodes of the second
object.
The cost-functional contains two constraint terms, one
penalizing the elastic deformation of edges between
neighboring nodes, the other penalizing mismatches between
the labels carried by pairs of mapped nodes. The best mapping
is thus a trade-off between structural integrity and feature
integrity.
We carried out numerical experiments on a public database
of 1200 16x16-pixel handwritten digits and compared the results
with traditional feed-forward layered network methods. Defining
our own graph-matching cost-functional as a "pseudo-metric" in
image space, we applied a simple nearest-neighbor decision
criterion and found significantly lower error rates. This
proved that designing an appropriate representation format
(here, 2-D relational) is often more important than fine-tuning
the parameters of a learning process.
I later created a new version of 2-D graph matching based on temporal
correlations and phase-locking in spiking neural networks
(Doursat et al. 1995).
|
References
von der Malsburg (1981) Correlation theory of brain function.
MPI for Biophys. Chemistry, Göttingen.
|
Publications
Doursat, R. (1991)
Contribution à l'étude des représentations dans le
système nerveux et dans les réseaux de neurones formels.
PhD Thesis, Université Pierre et Marie Curie (Paris 6).
|
|
|
An epigenetic development model of the nervous system
|
 |
The third part approached the issue of neural
representation from a more abstract and speculative viewpoint.
We wanted to address the compositionality
of cognitive processes and language, i.e., the faculty of assembling elementary
constituent features into complex representations.
Answering Fodor and Pylyshyn's (1988) influential criticism
about the lack of structured representations in neural networks,
we showed that compositionality
can arise from the simultaneous self-organization of
connectivity and activity in an initially random network.
|
Already apparent in invariant perceptual tasks, where objects
are categorized according to the relationships among their parts,
compositionality is particularly striking in language
and is also referred to as constituency. Language
is often described as a "building block" system, in which the
operative objects are symbols endowed with an internal
combinatorial structure. This structure allows elementary
symbols to be assembled in many different ways into
complex symbols, whose meaning is sensitive to their
internal arrangement. Again, chemistry provides a useful
metaphor if we compare symbols with molecules and symbolic
composition with the various possible reactions and products
that depend on the geometrical structure of molecules.
In this context, the issue of an appropriate format of
representation of mental entities is of particular importance
and our proposal is that the nervous system uses a higher-order
code to represent linguistic entities.
The goal of the present neural model was to show that
compositionality can arise from the gradual ontogenetic
development of the nervous system during the early stages of
synaptogenesis. By this, we adhered to Chomsky's conception
that language actually "grows" and matures in children's brain
like a limb or an organ. This claim
might sound suprising at first but is in accordance with
well-known observations and general principles of neural
development.
The visual system and many other cortical areas
display striking regularities in their connectivity, which
self-organize during fetal and postnatal development (with
or without input from external stimuli) and account for their
functional specialization. Similarly, it is postulated here
that the faculty of language (as opposed to any specific
language) is supported by specialized neural pathways that
develop through a feedback interaction between neuronal
activities and synaptic efficacies.
Starting from an initially disordered network with low
random activity, certain synaptic connections are gradually
selected and strengthened to the detriment of others. This
focusing of the connectivity is also accompanied by
a gradual increase and durability of correlated firing.
Connections and correlations reinforce each other
through heterosynaptic cooperation, while the global
stability of the network is maintained through a constraint
of competition.
On the whole, complex spatiotemporal patterns of spiking
activity spontaneously emerge within the network. These
patterns have been experimentally detected in mammalian
species and termed synfire chains (Abeles 1982).
In our claim, they constitute the elementary components or
building blocks of compositionality. Such patterns
have the required properties discussed above, i.e., an internal
combinatorial structure that allows them to assemble in multiple
ways, thereby opening the way to a virtually infinite hierarchy
of combinations.
For a more recent version of this study, see the
project.
|
References
Abeles (1982) Local cortical circuits. Springer.
Fodor & Pylyshyn ('88) Connectionism and cognitive architecture
Cognition 28: 3-71.
|
Publications
Doursat, R. (1991)
Contribution à l'étude des représentations dans le
système nerveux et dans les réseaux de neurones formels.
PhD Thesis, Université Pierre et Marie Curie (Paris 6).
|
|