Neural Computation

Biological systems are obviously able to process abstract information on (some part) of the states of neuronal and molecular networks. The constant theme in my research has been to understand how biology achieves such computation in the neural circuits of the cerebral cortex. One might consider that the principles of Turing and related concepts, which have provided powerful models for developing technological computing, could be easily extended to biology. In fact, however, those principles have provided relatively little insight into biological computation, probably because they assume that the machines themselves, as well as their initial program are granted as input (from an intelligent human designer). By contrast, the organization of the states and the task related transitions of the biological computational process arise through phylogenetic and ontogenetic self-configuration processes and execute autonomously without the intervention of an external source of intelligence. One result of the evolutionary search is the potent intelligence generated by the cortex, using neural circuits that do not resemble any man made technology, and whose algorithms remain obscure. My research program on neural computation has examined various aspects of this great puzzle. For example, I have studied the details of cortical circuitry (with Kevan Martin, Henry Kennedy and Colette DeHay); the computational properties and dynamical stability of abstract cortical circuits (with Ueli Rutishauser and Jean-Jaques Slotine); inference and other forms of processing in graphical networks of neurons (with my students Andreas Steimer and Dylan Muir); the principles of autonomous self-constructing systems in the context of cortical development (with my students Fred Zubler, Sabina Pfister, Andreas Hauri, Roman Bauer, and Gabi Michel, and also Michael Pfeiffer); and the application of this knowledge to neuromorphic systems (with Giacomo Indiveri, Shih-Chii Liu, and Tobi Delbruck).