Annual Report 2020

New Flatiron Institute Center Aims to Map the Brain’s Inner Workings

Center for Computational Neuroscience

We have long understood the equations that govern the physical world, thanks to the elegant symbiosis between theory and experiment among physicists. For instance, theoretical physicists provide explanations of how magnetism or gravity operates in different contexts, which experimentalists can then confirm, refine or refute. This iterative process has resulted in a robust understanding of the world’s physical forces.

By contrast — and not surprisingly — we have no comparable understanding of the brain. Neuroscientists have determined which brain regions are responsible for complex planning (primarily the frontal lobe) and which regions help us interpret emotions (the amygdala and hippocampus, among others). But despite outstanding experimental achievements, we do not yet have a total picture of how the parts of the brain function together. That’s because we still need to develop theories that can guide interpretation of the experimental results. 

Enter the newly launched Center for Computational Neuroscience (CCN) at the Flatiron Institute, its fifth discipline-based computational center. The CCN’s mandate is to stimulate a collaboration between experimental and theoretical neuroscientists. Headed by Eero Simoncelli, who will also maintain his appointment as Silver professor of neural science, mathematics, data science and psychology at New York University, the center began its work in fall 2020 and will take possession of its physical space by the end of summer 2021.

Simoncelli is a leading authority on how human brains process visual information, and he has worked to develop ever-more-powerful computer models that explain this capability throughout his career. He is a Howard Hughes Medical Institute investigator, a fellow of the Institute of Electrical and Electronic Engineers, and an Emmy Award recipient for the development of computational models to assess how viewers perceive the quality of visual images.

“After a thorough search, Eero Simoncelli seemed the perfect person to head the new unit,” says Simons Foundation chair Jim Simons. “The other Flatiron directors and I were thrilled that Eero accepted the position.”

One useful way to think of the brain is as an extremely powerful computer, but one with a unique set of operations. “Understanding how the brain works is a computational challenge,” Simoncelli says. “Sensory input and internal states are continuously combined and transformed to drive thoughts, memories and behaviors. Our goal at CCN is to help decode that complexity.” 

Simoncelli notes that, especially in recent years, neuroscience experiments have yielded massive datasets that can only be interpreted with powerful computers, such as those at the Flatiron. The CCN will provide fertile ground for theoreticians, computational scientists and experimentalists to collaborate in developing an improved understanding, through data analysis, of how the brain works.

A novel neural network for supervised learning. Unlike existing neural networks based on the backpropagation algorithm, this neural network respects biological constraints and could model computations in the human brain. Click the plus button to toggle additional elements in this image. Credit: S. Golkar, D. Lipshutz, Y. Bahroun, A. Sengupta and D. Chklovskii/NeurIPS 2020

Simoncelli studied physics as an undergraduate and draws inspiration from the time-honored symbiosis between theory and experiment that physicists have developed. Though it might seem evident that neuroscience should have forged a similar path long ago, Simoncelli notes that physicists had the advantage of being able to begin with easily observable forces and objects rather than interpreting the dynamical evolution of ephemeral, internal neural patterns. In addition, the computational power to record and analyze neural activity at scale, in the brains of both animals and humans, has existed for only a decade or two. 

The CCN will initially comprise two working groups. Simoncelli, building on his expertise in the neuroscience of vision, will lead the computational vision research group. Group members will use computational theory and models that help interpret how our brain’s sensory systems make predictions about the world around us. Their goal is to understand how these processes, at the level of neural populations and circuits, drive visual behavior.

Dmitri ‘Mitya’ Chklovskii leads the neural circuits and algorithms group, which was previously part of the Center for Computational Biology. Chklovskii, who is also a research associate professor of neuroscience and physiology at New York University, focuses on how the activities of individual neurons and neural circuits generate thought and behavior. “Our goal is to understand the specific function of every neuron,” Chklovskii says. This more mechanistic approach will complement Simoncelli’s more behavior-centric analysis.

The CCN will ultimately employ approximately 50 people, some of whom will have joint appointments at the CCN and surrounding academic institutions. Visiting scientists will also be part of the mix. Simoncelli expects that everyone associated with the center will come to the physical offices at least some of the time, as circumstances permit.

“In my experience, the pandemic has reaffirmed the value of in-person contact,” Simoncelli says. “If a project was already very well established when the pandemic began, then remote work proceeded reasonably well. But starting new projects from scratch has been much harder online than in person.”

Toward that end, Simoncelli — the son of an architect (and named after one) — has been heavily involved in the design of the CCN’s physical space. Most offices will accommodate a maximum of two people. Simoncelli feels that two people sharing an office can do focused work together in silence, whereas adding a third person tends to impede concentration. There will also be a diverse and distributed set of open work areas where people can discuss joint research at blackboards. And, importantly, Simoncelli’s diligent work to find a suitable pro-level espresso maker for the center will help to fuel the quantity and quality of research insights!

———

The launch of the CCN adds another player to the Simons Foundation’s array of efforts to understand the human brain. SFARI — the foundation’s first program — and its independent news publication Spectrum produce and disseminate new insights about the neuroscience of autism. SFARI’s research cohorts, such as the Simons Simplex Collection and the SPARK cohort, will continue to provide readily accessible raw data to scholars of autism and related fields the world over. The Simons Collaboration on the Global Brain (SCGB) seeks to understand the mechanisms of the neural activity that produces cognition. This program funds researchers across the globe, including Simoncelli. A related collaboration launched this year, the Simons Collaboration on Plasticity and the Aging Brain, concentrates on the study of the healthy aging human brain.  

The potential for synergy between these groups is significant. “It’s extremely exciting to have this new dimension to the study of the human brain in-house,” says Simons Foundation president Marilyn Simons. “Understanding how the brain works is one of the biggest intellectual challenges of our time.”