Research in
progress
There are many issues I am currently working
on, in collaboration with some colleagues
and
several enthusiastic students:
- A tutorial (and possibly new methods) for
fixing/reconstructing non-PSD matrices
- The integration of instance weighing
methods with feature weighing
methods
- The application of feature selection
methods to biomedical problems:
- in spectral data (project AIDTumour
for diagnosis of brain tumours using Magnetic Resonance
Spectroscopy data)
- in microarray gene expression data
- The similarity
neural network, a generalization of the RBF for
supervised learning problems and its development into a deep
neural network
- Development
and understanding of new kernels for kernel-based methods,
specifically:
- kernels for
categorical features
- kernels
able to handle missing data (without imputation)
- kernels for
real-valued vectorial data alternative to the RBF kernel
- kernels for
fuzzy features
- aggregation
of kernels beyond arithmetic means
- Machine
learning methods (and software)
for the Microbial
Source
Tracking problem
On the drawing board:
- Development of analytical solutions for the
regularization parameter in Tikhonov regularization
- Theoretical frameworks for similarity
measures
- Feature selection methods based on full multivariate binary distributions
If you feel interested in any of these topics,
please do contact me.