Algorithms and Methods

We are developing novel algorithms for signal and data analysis, pattern recognition, neural computation, advanced statistics, machine learning, and artificial intelligence.

In our algorithmic research in our group of Algorithms and Methods we concentrate on extending our core expertise on unsupervised machine learning methods. Prime examples are Self-Organizing Maps (SOM) and Independent Component Analysis (ICA).

ICA is further expanded to positive, nonlinear, or hierarchical models, as dictated by applications. We have recently developed a general framework for source separation algorithms called Denoising Source Separation (DSS). The main benefit of this DSS framework is that it allows for easy development of new source separation algorithms which are optimized for specific problems.

Further work on Bayesian Latent Variable Models will lead to efficient approximative techniques.

We are conducting research on distance metric learning and dependency exploration for statistical data mining. Supervised data exploration, by combining supervised metrics with unsupervised methods, is particularly useful in bioinformatics.

Independent Variable Group Analysis (IVGA) technique is developed for feature extraction.

Tree-structured extensions of the SOM are also studied.

Contact: Academy professor Erkki Oja, erkki.oja at hut.fi