preprints available

Fri Oct 2 09:55:19 EDT 1998

Dear connectionists,

attached please find abstracts and preprint locations of five manuscripts on:

optical recording of brain activity:

1. tissue optics simulations for depth-resolved optical recording
2. ICA analysis of optical recording data

biological modelling:

3. contrast adaptation, fast synaptic depression, and Infomax in visual
   cortical neurons
4. the role of non-linear lateral interactions in cortical map formation

ANN theory:

5. optimal hyperplane classifiers for pseudo-Euclidean and pairwise data

Comments are welcome!

Cheers

Klaus

-----------------------------------------------------------------------------

Prof. Dr. Klaus Obermayer         phone:  49-30-314-73442
FR2-1, NI, Informatik                     49-30-314-73120
Technische Universitaet Berlin    fax:    49-30-314-73121
Franklinstrasse 28/29             e-mail: oby at cs.tu-berlin.de
10587 Berlin, Germany             http://ni.cs.tu-berlin.de/

=============================================================================

Simulation of Scanning Laser Techniques for Optical Imaging of Blood-Related 
Intrinsic Signals

M. Stetter and K. Obermayer

Fachbereich Informatik, Technische Universitaet Berlin

Optical Imaging of intrinsic signals detects neural activation patterns by 
taking video images of the local activity-related changes in the
light intensity reflected from neural tissue (intrinsic signals). At red
light (605nm), these signals are mainly caused by local variations of the
tissue absorption following deoxygenation of blood.  In this work, we
characterize the image generation process during Optical Imaging by Monte
Carlo simulations of light propagation through a homogeneous model tissue
equipped with a local absorber. Conventional video-imaging and Scanning
Laser imaging are compared to each other. We find that, compared to video
imaging, Scanning Laser techniques drastically increase both the contrast
and the lateral resolution of optical recordings. Also, the maximum depth up
to which the signals can be detected, is increased by roughly a factor of 2
using Scanning Laser Optical Imaging. Further, the radial profile of the
diffuse reflectance pattern for each pixel is subject to changes which
correlate with the depth of the absorber within the tissue. We suggest a
detection geometry for the online measurement of these radial profiles,
which can be realized by modifying a standard Scanning Laser
Ophthalmoscope.

in: Journal of the Optical Society of America A, in press

available at: http://ni.cs.tu-berlin.de/publications/#journals

-----------------------------------------------------------------------------

Blind separation of spatial signal patterns from optical imaging records.

I. Schiessl^1, M. Stetter^1, J. Mayhew^2, S. Askew^2, N. McLoughlin^3,
J. Levitt^4, J. Lund^4, and K. Obermayer^5.

1 Fachbereich Informatik, Technische Universitaet Berlin
2 AIVRU, University of Sheffield
3 Department of Neurobiology, Harvard Medical School
4 Institute of Ophthalmology, UCL

Optical imaging of intrinsic signals measures two-dimensional neuronal activity
patterns by detecting small activity-related changes in the light reflectance
of neural tissue. We test, to what extend blind source separation methods, which
are based on the spatial independence of different signal components, are
suitable for the separation of these neural-activity related signal components
from nonspecific background variations of the light reflectance. Two ICA 
algorithms (Infomax and kurtosis optimization) and blind source separation by
extended spatial decorrelation are compared to each other with respect to their
robustness against sensor noise, and are applied to optical recordings from
macaque primary visual cortex. We find that extended spatial decorrelation is
superior to both the ICA algorithms and standard methods, because it explicitely
takes advantage of the spatial smoothness of the intrinsic signal components.

in: Proceedings of the ICA '99 conference, 1999 (accepted)

available at: http://ni.cs.tu-berlin.de/publications/#conference

-----------------------------------------------------------------------------

Influence of changing the synaptic transmitter release probability on contrast
adaptation of simple cells in the primary visual cortex.

P. Adorjan and K. Obermayer.

Fachbereich Informatik, Technische Universitaet Berlin

The contrast response function (CRF) of many neurons in the primary visual
cortex saturates, and shifts towards higher contrast values following prolonged
presentation of high contrast visual stimuli. Using a recurrent neural network
of excitatory spiking neurons with adapting synapses we show that both effects
could be explained by a fast and a slow component in the synaptic adaptation.
The fast component - a short term synaptic depression component - leads to a
saturation of the CRF and a phase advance in the cortical cells' response to
high contrast stimuli. The slow component is derived from an adaptation of the 
probability of the synaptic transmitter release, and changes such that the
mutual information between the input and the output of a cortical neuron is
maximal. This component - given by the infomax learning rule - explains
contrast adaptation of the averaged membrane potential (DC component) as well
as the surprising experimental results, that the stimulus modulated component
(F1 component) of a cortical cell's membrane potential adapt only weakly. Based
on our results we propose a new experimental method to estimate the strength of
the effective excitatory feedback to a cortical neuron, and we also suggest a 
relatively simple experimental test to justify our hypothesized synaptic
mechanism for contrast adaptation.

in: Advances in Neural Information Processing Systems NIPS 11, 1999 (accepted).

available at: http://ni.cs.tu-berlin.de/publications/#conference

-----------------------------------------------------------------------------

The role of lateral cortical competition in ocular dominance development.

C. Piepenbrock and K. Obermayer.

Fachbereich Informatik, Technische Universitaet Berlin

Lateral competition within a layer of neurons sharpens and localizes the
response to an input stimulus. Here, we investigate a model for the activity
dependent development of ocular dominance maps which allows to vary the degree
of lateral competition. For weak competition, it resembles a correlation-based
learning model and for strong competition, it becomes a self-organizing map.
Thus, in the regime of weak competition the receptive fields are shaped by the
second order statistics of the input patterns, whereas in the regime of strong
competition, the higher moments and ``features'' of the individual patterns
become important. When correlated localized stimuli from two eyes drive the
cortical development we find (i) that a topographic map and binocular, 
localized receptive fields emerge when the degree of competition exceeds a
critical value and (ii) that receptive fields exhibit eye dominance beyond a
second critical value. For anti-correlated activity between the eyes, the
second order statistics drive the system to develop ocular dominance even for
weak competition, but no topography emerges. Topography is established only
beyond a critical degree of competition.

in: Advances in Neural Information Processing Systems NIPS 11, 1999 (accepted).

available at: http://ni.cs.tu-berlin.de/publications/#conference

-----------------------------------------------------------------------------

Classification on pairwise proximity data.

T. Graepel, R. Herbrich, P. Bollmann-Sdorra, and K. Obermayer.

Fachbereich Informatik, Technische Universitaet Berlin

We investigate the problem of learning a classification task on data represented
in terms of their pairwise proximities. This representation does not refer to
an explicit feature representation of the data items and is thus more general
than the standard approach of using Euclidean feature vectors, from which
pairwise proximities can always be calculated. Our first approach is based on a
combined linear embedding and classification procedure resulting in an extension
of the Optimal Hyperplane algorithm to pseudo-Euclidean data. As an alternative
we present another approach based on a linear threshold model in the proximity
values themselves, which is optimized using Structural Risk Minimization. We 
show that prior knowledge about the problem can be incorporated by the choice
of distance measures and examine different metrics w.r.t. their generalization.
Finally, the algorithms are successfully applied to protein structure data and
to data from the cat's cerebral cortex. They show better performance than
K-nearest-neighbor classification.

in: Advances in Neural Information Processing Systems NIPS 11, 1999 (accepted).

available at: http://ni.cs.tu-berlin.de/publications/#conference