One of the most exciting and difficult challenges currently facing scientists is to understand how complex biological systems work.

The most complex of these systems - the brain - spans all levels of organization, from genes to channels to individual cells to networks of neurons to high brain functions. Computational neuroscience provides theoretical and computational tools that transcend many different levels of organization. Its approach is expanding into conventional neuroscience laboratories as the need for comprehensive analysis and interpretation of complex experimental data becomes increasingly difficult and important. The rapid growth of computational neuroscience and its future success depend on the establishment of highly productive collaborations between experimental labs and computational scientists. This multidisciplinary approach is at the very core of computational neuroscience.

Computational neuroscience has given especially important insight into higher visual activities. Studying how the brain learns to recognize and categorize visual objects is helping to understand the circuitry of the visual cortex as well as to design machines that learn from experience. In addition to mapping the pathways and physiological processes underlying visual learning, these studies will lead to improved diagnosis of various brain disorders and to the development of visual prostheses.

The McGovern Institute supports research on the higher brain functions of learning and visual recognition within a research group that integrates computational neuroscience with the mathematical theory of statistical learning and with its engineering applications. This approach enables researchers to use insights from studies of brain function to construct powerful learning theories that also have direct practical applications. For instance, McGovern faculty members have already demonstrated machine-learning systems in practical projects, including software that enables cars to distinguish pedestrians from other objects. These techniques can be extended to a variety of application areas, such as building machines that learn to diagnose specific cancer types by evaluating gene expression data or developing trainable, user-friendly man-machine interfaces.

The main goal of the computational neuroscience work in the McGovern Institute is to develop theories of higher brain functions that can be used as powerful tools to summarize existing data from system physiology and imaging, to interpret new results, and to plan new experiments in close collaboration with other experimental groups in the Institute. Ultimately the models will represent our growing understanding of complex brain functions and of how intelligence emerges from networks of neurons.

This research is also being done in the Center for Biological and Computational Learning.