Level 1 (instructions) |

**NB: This introduction is several years old.** The actual
thesis is here.

- Theory of modelling

Bayesian probability theory provides efficient tools for modelling. I have studied how the theory can be applied to neural networks, and the connections between Bayesian and information theoretic approach to modelling. - Unsupervised construction and learning of models
- Comparison with biological brain

The brain has had to solve how to construct a complex model and how to make it work efficiently. We can learn from the structure of the brain how to develop solutions for pattern recognition, and, on the other hand, good solutions might provide clues about the functions of different parts of the brain.

I have collected some links and references.

I'm first writing this presentation in Finnish, so there is a bit more material available in Finnish.

Harri Valpola <Harri.Valpola@hut.fi>