Research

Constructing an intelligent system begins with finding a way to describe the artificial neurons and how they can be connected. Then a basic structure, large and complex enough to solve the given task has to be defined.

Using Evolutionary Algorithms to develop cognitive systems

Evolution has proven capable of developing creatures with remarkable abilities. We use recent findings in evolutionary research to develop strategies of creating better networks. The basic idea is to generate a random network, and see how well it performs at a given task. Then we make a small random alteration of this network, and evaluate how well it performs. If it is better than the previous, we keep the new blueprint of a network, otherwise we discard the solution.

Cognitive technology in visual systems

Of the 20 billion neurons in the brain more than 40% of the information carried and processed involves vision. In the light of this fact we use existing image processing/filtering techniques trying to extract important features and characteristics reducing the data amount and feeding it to cognitive systems. Using neural simulators for image processing with millions of biological plausible neurons, the different image filtering techniques are tested with chip implementation in mind.