Friday, December 22nd, 2017

Looking Ahead: Spring 2018

by Luca Trevisan

The Brain and ComputationReal-Time Decision Making

Next semester, the Simons Institute for the Theory of Computing will host two programs, on theoretical neuroscience and on real-time decision making.

The program on The Brain and Computation, organized by Bruno Olshausen, Sophie Denève, Ila Fiete, Wolfgang Maass, Bartlett Mel, Christos Papadimitriou, Terry Sejnowski, and Santosh Vempala will bring to Berkeley an outstanding mix of neuroscientists and computer scientists, to investigate foundational questions at the intersection of the two disciplines.

One theme of the program will be the application of computational models and ideas to certain key questions in brain science, such as how to model knowledge representation in the brain, how to model the non-linear computations performed by neurons, how to interpret information about connectomics, and so on. Another theme will be the study of algorithms inspired by brain functions, and other applications of brain science to computer science. A third theme will be the study of computational models and theories that interpret and explain the large amount of data about brain anatomy and function that has recently become available.

The program will start with a week-long series of lectures on brain science, machine learning, and neural computing, designed to establish a common language among the participants. During the semester, three workshops will take place, on representation of information in the brain, on interpreting and modeling the available data on brain anatomy and function, and on computational theories in brain science.

The program comes at an ideal time, given the increasing availability of data about the brain and the emerging need of new theories and models, and we hope that it will have a long-lasting impact in forging new collaborations among neuroscientists and theoretical computer scientists.

The program on Real-Time Decision Making, organized by Richard Karp, Joshua Bloom, Steven Low, Evdokia Nikolova, and Balaji Prabhakar, is motivated by applications in science and engineering where: (1) a large amount of observational data is continuously generated, and (2) important decisions have to be made in real time based on the data observed up to that point. An example of such an application domain is astronomy: observatories generate a huge, continuous, amount of data, and when an anomalous event happens, one wants to quickly be able to detect it, and, if appropriate, change observation schedules to focus on the event. Other significant examples are transportation networks, earthquake warning systems and smart grids.

Although the above application domains are rather different from each other, they can be approached using a similar mathematical toolkit: traditional techniques from control theory and information theory; machine learning; and techniques closer to the core of theoretical computer science, such as streaming algorithms and distributed algorithms. Program organizers and participants include, among others, astronomers, traffic engineers, and experts in operations research, in distributed systems, in machine learning, and in algorithms.

The boot camp will have a great mix of overviews of applications and of tutorials on mathematical and algorithmic techniques. The three workshops will focus on applications, on the mathematics of social networks, and on computational challenges, respectively.

Like the brain science program, the real-time decision making program is exciting for its interdisciplinarity and for its potential to create new collaborations between scholars that would have not otherwise had a chance to learn from each other and to work together.

We look forward to a great semester.

Related Articles