The ComMIT graduate school offers a postgraduate study program focused in the computationally intensive methods of information technology, including neural computing and Bayesian modelling. The study program is offered by the Helsinki University of Technology and University of Helsinki.
The laboratories and institutions participating in the ComMIT are
Laboratory of Computational Engineering &
Research Centre for Computational Science and Engineering
Helsinki University of Technology
Laboratory of Computer and Information Science
Helsinki University of Technology
Biometry and Mathematical Methods of Information Technology
Rolf Nevanlinna Institute
University of Helsinki
Laboratory of Data Analysis,
Center for Mathematical and Computational Modeling,
University of Jyväskylä
In the research centre there are three research groups with overlapping research areas and topics:
Computational information technology
The research in the computational information technology group deals with problems where efficient numerical methods of computational engineering are combined with the problems of information processing. During recent years, computational approach in information processing has gained increasing importance, with progress in fields like computational intelligence and data mining.
In the group the research is focused in neural computing, Bayesian methods, and pattern recognition and modeling applications. Much of the emphasis in recent years has been in Bayesian modelling using Markov Chain Monte Carlo methods. We have developed and applied these methods for neural network models in several real world problems, and showed how many problematic issues of setting up and analysing a well-generalizing neural network can be solved.
In pattern recognition and machine vision we are studying 3D vision (reconstruction of shapes from images, camera calibration, stereo vision, inference of shape with Bayesian methods, etc.) and the design and assessment of classifier algorithms.
Cognitive science and technology
Research in cognitive science and technology focuses in the mechanisms of human audiovisual speech perception. We apply both psychophysics and modern brain research methods (EEG, MEG, fMRI) to reveal the critical features underlying unimodal speech perception. However, we are especially interested in how our brains integrate speech information obtained via both eyes and ears. To integrate the obtained results, we are developing models to explain the experimental findings. Recent projects include investigation of attentional influences on audiovisual speech perception, the effect of audio signal-to-noise ratio on audiovisual integration, and the robustness of the most prominent model for audiovisual integration, the Fuzzy Logical Model of Perception. We are also developing the Artificial Person, a dynamic model of human speech and emotion gestures. This is a computer model producing audiovisual speech and emotion gestures. Our emphasis is in using the Artificial Person in a dialogue system. For this aim, we are studying and modelling human social interaction. A successful candidate would participate in one of the following projects:
Computational materials research
In computational materials research the focus is on materials properties and micro- and nano-scale systems and their applications to new information technologies. Recent topics include structural properties of solids and soft biological materials, electronic and optoelectronic materials, physical layer of optical networks, microelectromechanical systems, and quantum information processing. In addition, computational algorithms, graphical visualisation and animation, and parallel computing methods are being developed for these models and high performance computations.
Neural Networks Research Centre (NNRC)
The Neural Networks Research Centre (NNRC) and the Laboratory of Computer and Information Science have long scientific traditions dating back to the 1960s. The main research topic for the last 20 years has been the study of neural computing, especially methods for unsupervised learning and their applications. The best-known innovation is the Self-Organizing Map, developed by Academician Teuvo Kohonen. It has been recently applied to industrial processes, telecommunications, biomedical problems, as well as to text and image document search in large databases. Independent Component Analysis, a statistical signal and data processing technique, has recently been under intensive research. The key applications studied in the projects include the analysis of brain signals through MEG, finding hidden factors in financial data, separating telecommunication signals and finding new methods for signal preprocessing. The NNRC, whose head is Academy Professor Erkki Oja, is one of the national Centres of Excellence.
From Data to Knowledge Research Unit
The Laboratory of Computer and Information Science is also partly hosting another Centre of Excellence, From Data to Knowledge Research Unit, in which Professor Heikki Mannila is a partner. The research is concerned with the development of computer methods for the processing of large and complex data materials. These methods will help people to find useful information from the data they have discovered. This is a multidisciplinary effort: the research teams working under the unit's umbrella combine know-how in the fields of efficient algorithms, statistical methods, database and machine learning techniques as well as application sciences. The key tools applied are combinatorial pattern matching and data mining. The combination of these two approaches represents an internationally unique feature. The results have practical application in molecular biology and bioinformatics, process industry, datacommunications, ecology and language technology.
Rolf Nevanlinna Institute is a research institute of Mathematics, Computer Science and Statistics located at the Faculty of Science of University of Helsinki. The main tasks of RNI are research and doctoral training. The research carried out at the Institute is highly multidisciplinary and it is characterized by active collaboration with many academic research groups, government research organizations, as well as private industry. About half of the staff of almost 40 are Ph.D. students and in recent years 3 - 4 doctoral theses have been produced annually.
The research staff is organized into three research divisions: Biometry, Mathematical Methods of Information Technology, and Mathematical Theory and Applications of Electromagnetic Fields. ComMIT will include the divisions of Biometry and Mathematical Methods of Information Technology.
Division of Biometry
The goal of the Division of Biometry is to carry out original methodological and applied research, with an emphasis on concrete scientific problems coming from biology, medicine, and public health. Much of the emphasis in recent years has been on the development and application of Bayesian methodology based on large hierarchically structured models, and on the application of algorithmic Markov chain Monte Carlo simulation techniques in the numerical computations. The statistical genetics group of the Division is a partner of the "Centre of Population Genetic Analyses", which enjoys the status of a Centre of Excellence of the Academy of Finland during the period 2002 - 2007. New graduate students are recruited e.g. to the following research projects:
Division of Mathematical Methods of Information Technology
The main area of research of the Division of Mathematical Methods of Information Technology has been the development and application of statistical and computational tools for the analysis on complex data. Examples of such tools are nonparamteric estimation of density and regression functions as well as artificial neural networks. Applications have centered on data-analysis problems in many areas of science and technology. Often the applications have involved statistical pattern recognition. New Ph.D. students are sought to two ongoing projects:
Laboratory of Data Analysis is a multidiciplinary research unit which brings together statistical and computer science methodogy for analysis of statistical data. The following research areas are included: pattern recognition and image analysis, Bayesian statistics, computationally intensive statistical methods (including MCMC), spatial statistics, stochastic simulation, and robust multivariate methods. Current areas of application are modelling of industry processes, materials science, forestry, ecology and environmental sciences, public health and signal processing.