No subject

[Klaus Obermayer]

Dear Connectionists,

attached please find abstracts and preprint-locations of four papers about:

1. the application of Gold et al.'s (1995) matching method to the measurement
of flow fields in fluid dynamics

2. Bayesian transduction

3. ICA and optical recording of brain activity

4. the role of cortical competition in visual cortical information processing

Cheers

Klaus

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

A new particle tracking algorithm based on deterministic annealing and
alternative distance measures

M. Stellmacher and K. Obermayer

CS Department, Technical University of Berlin, Germany

We describe a new particle tracking algorithm for the interrogation of double
frame single exposure data which is obtained with particle image velocimetry.
The new procedure is based on an algorithm which has recently been proposed by
Gold et al. (1995) for solving point matching problems in statistical pattern
recognition. For a given interrogation window, the algorithm simultaneously
extracts: (1) the correct correspondences between particles in both frames
and (2) an estimate of the local flow-field parameters. Contrary to previous
methods, the algorithm determines not only the local velocity, but other local
components of the flow field, for example rotation and shear. This makes the
new interrogation method superior to standard methods in particular in regions
with high velocity gradients (e.g. vortices or shear flows). We perform
benchmarks with three standard particle image velocimetry (PIV) and particle
tracking velocimetry (PTV) methods: cross-correlation, nearest neighbour search,
and image relaxation. We show that the new algorithm requires less particles per
interrogation window than cross-correlation and allows for much higher particle
densities than the other PTV methods. Consequently, one may obtain the velocity
field at high spatial resolution even in regions of very fast flows. Finally,
we find that the new algorithm is more robust against out-of-plane noise than
previously proposed methods.

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

to appear in: Experiments in Fluids

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

Bayesian Transduction

Graepel, R. Herbrich, and K. Obermayer

CS Department, Technical University of Berlin, Germany

Transduction is an inference principle that takes a training sample and aims at
estimating the values of a function at given points contained in the so-called
working sample. Hence, transduction is a less ambitious task than induction
which aims at inferring a functional dependency on the whole of input space. As
a consequence, however, transduction provides a confidence measure on single
predictions rather than classifiers, a feature particularly important for
risk-sensitive applications. We consider the case of binary classification by
linear discriminant functions (perceptrons) in kernel space. From the transductive
point of view, the infinite number of perceptrons is boiled down to a finite
number of equivalence classes on the working sample each of which corresponds
to a polyhedron in parameter space. In the Bayesian spirit the posteriori
probability of a labelling of the working sample is determined as the ratio
between the volume of the corresponding polyhedron and the volume of version
space. Then the maximum posteriori scheme recommends to choose the labelling
of maximum volume. We suggest to sample version space by an ergodic billiard
in kernel space. Experimental results on real world data indicate that Bayesian
Transduction compares favourably to the well-known Support Vector Machine, in
particular if the posteriori probability of labellings is used as a confidence
measure to exclude test points of low confidence.

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

to be presented at NIPS 1999

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

Application of blind separation of sources to optical recording of brain activity

H. Schner^1, M. Stetter^1, I. Schiel^1, J. Mayhew^2, J. Lund^3, N. McLoughlin^3,
and K. Obermayer^1

1: CS Department, Technical University of Berlin, Germany
2: AIVRU, University of Sheffield, UK
3: Institute of Ophthalmology, UCL, UK

In the analysis of data recorded by optical imaging from intrinsic signals
(measurement of changes of light reflectance from cortical tissue) the removal of
noise and artifacts such as blood vessel patterns is a serious problem. Often
bandpass filtering is used, but the underlying assumption that a spatial frequency
exists, which separates the mapping component from other components (especially
the global signal), is questionable. Here we propose alternative ways of
processing optical imaging data, using blind source separation techniques based on
the spatial decorrelation of the data. We first perform benchmarks on artificial
data in order to select the way of processing, which is most robust with respect
to sensor noise. We then apply it to recordings of optical imaging experiments
from macaque primary visual cortex. We show that our BSS technique is able to
extract ocular dominance and orientation preference maps from single condition
stacks, for data, where standard post-processing procedures fail. Artifacts,
especially blood vessel patterns, can often be completely removed from the maps.
In summary, our method for blind source separation using extended spatial
decorrelation is a superior technique for the analysis of optical recording data.

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

to be presented at NIPS 1999

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

Recurrent cortical competition: Strengthen or weaken?

P. Adorjn, L. Schwabe, C. Piepenbrock, and K. Obermayer

CS Department, Technical University of Berlin, Germany

We investigate the short term dynamics of recurrent competition and neural
activity in the primary visual cortex in terms of information processing and in
the context of orientation selectivity. We propose that after stimulus onset,
the strength of the recurrent excitation decreases due to fast synaptic
depression. As a consequence, the network is shifted from an initially highly
nonlinear to a more linear operating regime. Sharp orientation tuning is
established in the first highly competitive phase. In the second and less
competitive phase, precise signaling of multiple orientations and long range
modulation, e.g., by intra- and inter-areal connections becomes possible
(surround effects). Thus the network first extracts the salient features from
the stimulus, and then starts to process the details. We show that this signal
processing strategy is optimal if the neurons have limited bandwidth and their
objective is to transmit the maximum amount of information in any time interval
beginning with the stimulus onset.

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

to be presented at NIPS 1999

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

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/