Machine Learning and Statistical Physics
Initiative for the Theoretical Sciences, The Graduate Center, CUNY
November 13 - 15, 2018
Recent years have seen dramatic progress in machine learning and artificial intelligence, driven in large part by the success of deep neural networks. These systems are constructed out of very many simple units that collectively give rise to remarkable performance on diverse tasks from object recognition to playing games such as Go.
However, how and when these methods work is largely not understood. Since their remarkable performance is an emergent property of many interacting units, the framework of statistical physics should provide fundamental insights into these systems.
This workshop brings together a number of researchers taking a statistical physics approach to machine learning with the intention of using insights from physics to understand learning systems. While there is a long history of productive interaction between statistical physicists working in spin glasses/disordered systems and machine learning practitioners, here we intend to broaden the scope to other applications of statistical physics ideas as well.
We hope to emerge from this workshop with a clear set of directions for promising future research of machine learning.
Tankut Can (GC)
Sarang Gopalakrishnan (GC/CSI)
Vadim Oganesyan (GC/CSI)
David Schwab (GC)