Computational information technology 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**. Computational intelligence
comprises methods such as neural networks, probabilistic methods,
fuzzy systems and evolutionary computing, where iterative numerical
methods are used for solving complex problems, involving large
amounts of measured data and only uncertain prior knowledge about the
solutions or modeled phenomena. A central element is inference in
novel situations, based on observed data, that is, inductive learning
systems. Similar capabilities for knowledge discovery are needed in
data mining, where useful pieces of information are sought from large,
possibly unorganized, databases.

In the laboratory the research in computational information technology is focused in neural computing, Bayesian methods, and pattern recognition and modeling applications. A good example of fields in information processing, in which computational methods have large impact, is the Bayesian approach, where the estimation of a specific set of model parameters is replaced by constructing the probability distribution of the model parameters and integrating over it during the use of the model. This integration requires efficient numerical methods, such as Markov chain Monte Carlo methods. The numerical solution also makes it possible to avoid some standard assumptions (such as linearity and/or gaussianity), that may be bad in practice. For an overview of the approach and our results, see review paper [45].

In the following chapters we give short descriptions of the research projects in the computational information technology group.

- Bayesian Methods for Neural Networks
- Probability Density Model for the Self-Organizing Map
- Neural Networks in Electrical Impedance Tomography
- Learning Scene and Object Analysis
- Building Spatial Choice Models from Aggregate Data
- Compound Analysis by combining Infrared and Mass Spectroscopy
- The Adaptive Brain Interfaces
- Internet Services Modelling
- Multiple View Geometry in Computer Vision
- Computer Vision for Electron Tomography
- Statistical Analysis and Modeling of Asset Returns
- Stochastic Simulation of Traffic Flow
- Temporal Difference Learning Applied to a High Performance Game-Playing Program