Physics-Inspired Computation

Our research group focuses on Physics-Inspired Computation (PIC), particularly within the framework of Probabilistic Computing, aiming to pioneer non-Von Neumann architectures. We utilize novel non-volatile electronic devices (such as RRAM and MTJ) to develop innovative circuitry and specialized hardware solvers. The core of our work involves creating and implementing three fundamental computational primitives: the Probabilistic-bit (p-bit), the Gaussian-bit (g-bit), and the Analog-bit (a-bit).

By harnessing these primitives, we efficiently tackle a wide spectrum of computational challenges, including NP-hard combinatorial optimization problems like Integer Factorization, Boolean Satisfiability (SAT), the Traveling Salesman Problem (TSP), and complex bio-molecular simulations such as RNA/Protein Folding and Molecular Docking. Furthermore, we are addressing mixed and continuous optimization problems, including 0-1 Integer Programs and the acceleration of Deep Learning model training and hyperparameter tuning. 

 

Our goal is to achieve a deep co-optimization among emerging nanodevices, new computing architecture, algorithm design and system-level hardware implementations, ultimately advancing probabilistic computing systems.