- Feature Story
BCS Welcomes New Associate Professor
Alexander “Sasha” Rakhlin studies models of learning and decision-making
For Prof. Alexander “Sasha” Rakhlin, mathematics is so much more than numbers and equations on a page. Growing up in Russia, he attended one of the top math schools in the country, but it wasn’t until he studied computer science and statistics as an undergraduate at Cornell University when he realized its true potential to have a direct impact on the world.
“I became fascinated with the idea that one can mathematically model the process of learning from data or from experience in the outside world, write a computer program based on the model, and observe the learning process in front of your eyes” says Rakhlin. “That led to my interest in the brain. I was interested in the mechanisms of how learning can be implemented, and in understanding it in computational terms so we can build models to replicate that phenomena in man-made systems.”
Rakhlin’s interest in the brain and statistical expertise drew him to MIT, where he completed his graduate studies with Prof. Tomaso Poggio, who is currently the director of the Center for Brains, Minds and Machines (CBMM), founded in 2005. Rakhlin joined the predecessor of CBMM, the Center for Biological & Computational Learning, which was founded in 1992 and combined mathematics, engineering, and neuroscience to better understand biological and artificial intelligence. According to Rakhlin, statistics is an important tool in computational neuroscience, particularly when trying to understand how learning works in an intelligent system.
“Statistics plays a huge role in neural research because it provides the necessary tools for extracting signal from noise, for finding structure in the data. In fact, a lot of progress in areas like machine learning came in a large part from theoreticians suggesting new methods based on their analyses,” says Rakhlin. “In addition to classical statistical approaches, I think that our research should focus on building models that learn sequentially from data since that is how a person learns. We can see what works, what doesn’t work, and update the models as we learn from experience.”
Post-graduate school, Rakhlin honed his research program at the University of California at Berkeley for his postdoctoral training and as a faculty member in the Department of Statistics at the University of Pennsylvania. Today, his research focuses on models of learning and decision-making. He has devoted a large part of his research to developing the theoretical and algorithmic foundations of online (sequential) learning in dynamic environments. He began this research by studying extensions of classical statistical techniques for independent data and later switched to the more realistic and challenging problem of modeling dependent data.
His work has resulted in a range of new methods for making online predictions that show excellent performance on real-world data. Most notably, his active research in online learning is pointed at one of the most important open questions at the intersection of the brain sciences and artificial intelligence — how does the brain learn so efficiently relative to current machines? Rakhlin is currently setting up his teaching program here, where he hopes to focus on the statistical analysis of neural data, machine learning, and more.
With the Department’s commitment to intelligence research through CBMM and the newly-formed MIT Intelligence Quest initiative, Rakhlin looks forward to expanding and strengthening the connections between BCS and other MIT centers, such as IDSS, CSAIL and more.
“For me, the most exciting thing is how math has a real impact on the world. IDSS will coach students in mathematical and applied statistics and ensure that it has real-world applications, including economics, computer science, and of course, brain and cognitive sciences,” says Rakhlin. “I see a lot of opportunity for natural synergy between BCS and IDSS, particularly when incorporating computational approaches in neuroscience research, and I’m excited to help build that bridge to the benefit of the wider MIT community.”
Learn more about Prof. Rakhlin’s research.