The Statistical Physics of of machine
learning has re-gained considerable attention in the
Tools and numerical methods for the systematic analysis of
processes provide, for instance, typical
learning curves in clear-cut model situations.
Besides analysing equilibrium situations,
the framework allows for the mathematical description of the
learning dynamics in multi-layered neural network architectures
and other systems.
Our current efforts concern the modelling of
learning in the presence of different forms of concept drift. Furthermore we extend earlier studies of multi-layered
neural networks can be extended to techniques
that have recently re-gained popularity in the context of so-called
Deep Learning approaches.
Deep theoretical analysis provides valuable insights, helps to optimize existing training schemes and
facilitates the development
of entirely new approaches.
Very early, the statistical physics approach has
led to, for instance, the formulation of the
so-called AdaTron algorithm
which has gained significant popularity
in the context of Support Vector Machines:
More recently, the use of adaptive distance measures
and relevance learning in prototype-based classifiers
has been in the center of my interest. The method of Matrix Relevance Learning
is based on the use of
adaptive metrics in Learning Vector
Quantization (LVQ). It has been shown to improve
classification performance significantly in many cases
and to provide, at the same time, deep insight into the problem
at hand. The basic formalism was communicated in:
top The application of newly developed methods and improved
algorithms constitutes an integral part of our
research. Many relevant theoretical
questions and methodological challenges can only be
in the context of real world data and practical applications.
To this end, it is vital to establish and
maintain intense collaborations with the corresponding domain
In the following, only a few selected examples of on-going
projects are highlighted in order
to illustrate the diversity of the considered problems.
We have applied machine learning successfully in the
context of various medical diagnosis
problems. A long
standing collaboration with the U.of Birmingham,
Metabolism and Systems Research,
constitutes a particular
successful example. Steroid metabolomics data is analysed
for the diagnosis and monitoring of tumors
and other disorders of the adrenal glands.
A recent, prospective study of our efforts paves the path for
using the method in clinical practice:
We have applied a variety of supervised and unsupervised machine
learning techniques in
bioinformatics and systems biology. Examples include the analysis of genomics and proteomics
data, and risk evaluation in cancer based on gene expression. A
recent publication provides evidence
for tissue- and stage-specific
composition of the ribosome:
In a collaboration with
neuroimaging and neurology at the UMCG Groningen
we analyse FDG-PET brain scans and motion sensor data.
The ultimate goal is the
early diagnosis of neurodegenerative disorders.
In this context, the combination of
dimension reduction with the intuitive, interpretable GMLVQ has proven useful:
Crop plant diseases are in the
focus of a collaboration with
AI Lab, Makerere University
in Kampala and
NaCRRI, the National
Crops Resources Research Institute of Uganda.
Image and spectral data are analysed for the early diagnosis
and discrimination of
viral diseases in Cassava. A recent publication:
Activities embedded in the H2020 ITN
SUNDIAL as well as collaborations
with individual researchers concern
the analysis of (radio-) astronomical
data. One example application is the
types as presented in: