Boris
Hanin
Head of Scientific Advisory Board
Dr. Boris Hanin is Head of the Scientific Advisory Board at Logical Intelligence and an Associate Professor in Operations Research and Financial Engineering at Princeton University, where he is also Associated Faculty in the Program in Applied and Computational Mathematics. His research is in deep learning theory, probability, and spectral asymptotics. Its central aim is a rigorous mathematical understanding of why neural networks work and how they behave at scale.
His work has produced foundational results on how network architecture, width, and depth shape training. These include a characterization of wide neural networks as Gaussian processes, finite-size corrections to the neural tangent kernel, and principled methods for transferring hyperparameters across model scales in transformers, mixture-of-experts, and graph networks. He has also studied the scaling behavior of compound AI systems with collaborators at Stanford and Berkeley. His work appears at NeurIPS, ICML, and ICLR and in journals including PNAS, Physical Review Letters, Acta Numerica, Communications in Mathematical Physics, and the Journal of Machine Learning Research. He also contributed research to The Principles of Deep Learning Theory (Cambridge University Press).
Before Princeton, Dr. Hanin was an Assistant Professor of Mathematics at Texas A&M University and an NSF Postdoctoral Fellow in Mathematics at MIT. His research is supported by a 2024 Sloan Fellowship in Mathematics, an NSF CAREER award, and a DARPA grant. He serves as Associate Editor of Mathematics of Operations Research, Pure and Applied Analysis, and Advances in Theoretical and Mathematical Physics.
A leading educator in machine learning theory, Dr. Hanin organizes the Princeton Machine Learning Theory Summer School. He has taught short courses at Les Houches, Oxford, the University of Chicago, and the Lake Como School. He holds a PhD in Mathematics from Northwestern University, where he studied under Steve Zelditch.
Dr. Boris Hanin is Head of the Scientific Advisory Board at Logical Intelligence and an Associate Professor in Operations Research and Financial Engineering at Princeton University, where he is also Associated Faculty in the Program in Applied and Computational Mathematics. His research is in deep learning theory, probability, and spectral asymptotics. Its central aim is a rigorous mathematical understanding of why neural networks work and how they behave at scale.
His work has produced foundational results on how network architecture, width, and depth shape training. These include a characterization of wide neural networks as Gaussian processes, finite-size corrections to the neural tangent kernel, and principled methods for transferring hyperparameters across model scales in transformers, mixture-of-experts, and graph networks. He has also studied the scaling behavior of compound AI systems with collaborators at Stanford and Berkeley. His work appears at NeurIPS, ICML, and ICLR and in journals including PNAS, Physical Review Letters, Acta Numerica, Communications in Mathematical Physics, and the Journal of Machine Learning Research. He also contributed research to The Principles of Deep Learning Theory (Cambridge University Press).
Before Princeton, Dr. Hanin was an Assistant Professor of Mathematics at Texas A&M University and an NSF Postdoctoral Fellow in Mathematics at MIT. His research is supported by a 2024 Sloan Fellowship in Mathematics, an NSF CAREER award, and a DARPA grant. He serves as Associate Editor of Mathematics of Operations Research, Pure and Applied Analysis, and Advances in Theoretical and Mathematical Physics.
A leading educator in machine learning theory, Dr. Hanin organizes the Princeton Machine Learning Theory Summer School. He has taught short courses at Les Houches, Oxford, the University of Chicago, and the Lake Como School. He holds a PhD in Mathematics from Northwestern University, where he studied under Steve Zelditch.
▧