Research Interests:


My long-term scientific goal is to solve fundamental problems in AI, particularly the unsupervised learning problem. I finished my PhD at MIT in theoretical physics with Patrick Lee, where I worked on quantum field theories for strongly correlated systems. I did my postdoc at the Salk Institute with Terry Sejnowski, where I developed a novel framework for image modeling. I recently moved to UC Berkeley as a Researcher at the Redwood Center for Theoretical Neuroscience. I’ve recently led and worked on projects on generative adversarial networks, score matching networks, and higher-order Hopfield networks.


Selected Works:


Mehrjou, A., Schölkopf, B. & Saremi, S. "Annealed Generative Adversarial Networks", arXiv:1705.07505 (2017).


Saremi, S. & Sejnowski, T. J. “Correlated percolation, fractal structures, and scale-invariant distribution of clusters in natural images”, IEEE Transactions on Pattern Analysis and Machine Intelligence (2016).


Saremi, S. & Sejnowski, T. J. “Hierarchical model of natural images and the origin of scale invariance”,  Proceedings of the National Academy of Sciences (2013).


Saremi, S. “RKKY in half-filled bipartite lattices: Graphene as an example”, Physical Review B (2007).


Selected Talks:


CIFAR Learning in Machines and Brains, Vancouver, Canada, April 2018 –– Deep Energy Estimator Networks.


CIFAR Learning in Machines and Brains, Long Beach, CA, December 2017 –– Generative score-matching networks.


CIFAR Learning in Machines and Brains, Paris, France, May 2017 –– Annealed generative adversarial networks.


CIFAR Learning in Machines and Brains, Barcelona, Spain, December 2016 –– Learning 4th-order Boltzmann machine in natural images.


Simons Foundation, New York, NY, October 2016  ––  The problem of many scales of length in natural images.


The Abdus Salam International Centre for Theoretical Physics (ICTP), Trieste, Italy, March 2016 –– Problems in physics, machine learning, and neuroscience with many scales of length.


Redwood Center for Theoretical Neuroscience, UC Berkeley, Berkeley, CA, April 2015 –– Correlated percolation, fractal dimensions, and scale-invariant distribution of clusters in natural images.


IBM Thomas J. Watson Research Center, Yorktown Heights, NY, May 2013 –– Hierarchical model of natural images and the origin of scale invariance.