Gatsby Program in Theoretical Neuroscience
Edmond and Lily Safra Center for Brain Sciences
Hebrew University of Jerusalem
The last two decades have witnessed the development and rapid growth of theoretical and computational neuroscience as a multi-disciplinary field at the forefront of brain science. The goals of this field span a broad range of activities, including the development and application of new tools for analysis of experimental data about brain structure and function, mathematical modeling and computer simulations of the activity of neurons, synapses and neuronal networks, development of theories of coding and decoding of information in neuronal circuits, exploration of neural theories of cognitive functions, and modeling mechanisms underlying brain dysfunction in neurological and psychiatric disorders. For reviews of the field see articles in Current Opinion in Neurobiology Volume 25, Pages 1-236 (April 2014).
The field strives to develop close collaboration with experimentalists to ensure that models are constrained by real data about brain and behavior, and to guide hypothesis driven experimental paradigms. In addition, the field is enriched by fruitful cross -interactions with theoretical physics, dynamics, statistics and computer science. With the recent emergence of powerful neurotechnologies and the transformation of the neuroscience into a data rich field, the need for theories, model and computational techniques is rapidly growing-and so are the challenges.
Recent revolutionary advances in AI are based largely on Deep Learning (DL) algorithms that tune the connections in neurally inspired deep neural networks to perform desired tasks on their inputs. DL algorithms have been used successfully in pattern classification and recognition, scene understanding, speech recognition and text translation, autonomous navigation, planning and acting and more. Other deep networks are trained to function as generative models, learning from examples to produce data of desired high dimensional, complex statistical structures.
DL provides a host of new powerful data analysis and modeling tools in vast areas of neuroscience, from reconstruction of circuits from EM imaging (connectomics) to modeling the layered structure of retina in the fly brain or the complex input-output relations of single neurons. Despite their resemblance, there exists fundamental gaps between artificial and brain systems both in terms of their architecture, dynamics and learning algorithms as well as with respect to their performance.
The Gatsby Program in Theoretical Neuroscience at ELSC, in collaboration with the Machine Learning and AI groups at HUJI CS at HUJI school of Engineering is in a unique position to make substantial advances at the interface between Deep Learning and real neuronal information processing.
The Gatsby Charitable Foundation’s pioneering investment in theoretical and computational neuroscience began in the 1990s with the establishment of the Gatsby Computational Neuroscience Unit at University College London. Subsequently, the Foundation has expanded its investment by creating a Tri-Center allegiance, between the Gatsby Unit at UCL, London, The Center for Theoretical Neuroscience, at Columbia University, New York, and the Gatsby Program in Theoretical Neuroscience, at The Edmond and Lily Safra Center for Brain Science (ELSC), at The Hebrew University of Jerusalem (formerly ‘The Interdisciplinary Center for Neural Computation’ (ICNC)). Over the last ten years, the three centers have organized joint meetings, programs of research, and student exchanges. It has also sponsored close collaborations with experimentalists at each location and across the three universities.
About the Program
The Gatsby Program in Theoretical Neuroscience at ELSC supports research in theoretical and computational neuroscience at the Hebrew University. The Program provides doctoral and postdoctoral fellowships, and sponsors conferences and visitors. Preference is given to students and postdocs who perform collaborative research mentored jointly by theorists and experimentalists. A priority area for the Program is sponsoring activities that strengthen the ties between this Program and the two partner centers: The Gatsby Unit for Computational Neuroscience (UCL, London) and The Center for Theoretical Neuroscience (Columbia University, NY). These activities include joint meetings, exchange of students, postdocs and faculty, and collaborative research.
Gatsby Fellows Program
In addition to regular postdoctoral fellowships, the program provides special fellowships for outstanding junior researchers at their senior postdoctoral level working on theoretical and computational neuroscience at the Hebrew University. Fellows are admitted to the Program based on their credentials and are free to choose their research agenda and their host lab in the course of their first semester in residence. In addition to upgraded salaries and housing support, Fellows receive modest research funds aimed primarily for travel to other labs or conferences. Fellows are expected to actively participate in initiatives that strengthen Tri-Center interactions.