Our research centers around mathematical modeling and computer simulation of biological nervous systems, the brain in particular. The primary aim is to improve our understanding of the principles underlying their information processing capabilities. The models used extend from those incorporating biological details of cellular processes, neurons and neuronal networks to abstract connectionist type models. Our field of research, theoretical and computational neuroscience, is quite young but has in recent years achieved a broad acceptance.
Being a computer science group, it is natural for CBN to emphasize also the engineering and applications aspects of understanding biological information processing, namely the use of this knowledge in the design of advanced technical systems. The nervous system is unique among biological entities since, like the computer, it is specifically aimed for information processing. The human brain still outperforms today's supercomputers by several orders of magnitude in terms of processing and memory capacity as well as in compactness, robustness, and power dissipation. No technical design reaches anywhere near its performance in tasks like vision, speech and language understanding, or motor control. Not surprisingly, the brain is often seen as a model for future man-made learning autonomous systems. Despite considerable progress in recent decades, the working principles of the brain remain mysterious. We regard it as our major research challenge to find out how such a performance is possible and to contribute to uncover the underlying information processing principles.
The CBN research group was formed in 1987 (called SANS - Studies of Artifical Nervous Systems). The group benefits from a long-standing and close collaboration with neurophysiologists at the Dept. of Neuroscience, Karolinska Institutet (KI, Prof. S. Grillner). In later years, collaboration has been further extended to other experimental groups in neuroscience (Prof. M. Ingvar, Dept. of Clinical Neurophysiology at KI; Prof. H. Forssberg, Dept. of Women and Child Health, KI) and neuropsychology (Prof. L.-G. Nilsson, Dept. of Psychology, Stockholm university). The group is partner of Stockholm Brain Institute (SBI) and leads its computational modeling methodology platform.
The major components of the brain are the sensory-perceptual systems, the behavior-motor systems, and the sub-cortical motivational and emotional systems. Throughout, plasticity, self-organization, and learning are fundamental properties. Over the years, the CBN group has conducted research relating to all these components and their function including associative memory and stimulus-response learning, generation and control of locomotion, behavior selection and reinforcement learning.
A man-made autonomous system, be it a robot searching for mines or an electronic agent negotiating for bandwidth on the Internet, faces much the same challenges as a living animal. For both, it is crucial to get a quick grasp on the situation at hand, to identify alternative actions, to decide which action to execute, and, finally, to do so before it is too late. The execution itself must be accurate and effective. The structure of the cerebral cortex is surprisingly homogenous with small differences between the different systems and areas, whether these are mainly engaged in sensory-perceptual or motor-action tasks. This indicates that general principles are widely used, allowing for efficient global coherence and integration. The idea of characterizing these principles and demonstrating how they can be tranferred into future autonomous learning systems penetrates most of our research.
CBN projects deal with: