Kerem Camsari
University of California-Santa Barbara, CA·Engineering·Early-career
Modern AI depends on huge amounts of energy, while ordinary transistors are no longer getting much more efficient. Some key computing tasks, such as checking whether a neural network is safe, are especially hard for conventional computers.
- Fields
Engineering
Computer Science
Physics
- Topics
probabilistic computing
p-bits
p-computers
- Grants
- 2 known grants, through Dec 2027
What they work on
- Improve the energy efficiency of probabilistic bits (p-bits) by using voltage-controlled magnetic anisotropy (VCMA) in spintronic devices.
- Design array-level spintronic p-computers, study how p-bits interact with feedback circuits, and emulate systems of more than 1,000 p-bits on FPGAs.
- Reframe exact deep neural network verification as an Ising model problem and solve it with hybrid classical/probabilistic computers and simulated quantum annealing.
- Build a large-scale probabilistic computer combining 10,000 FPGA-based digital p-bits with 100 stochastic magnetic tunnel junction p-bits that supply true randomness.
- Apply sparse, asynchronous p-computers to probabilistic and quantum machine learning algorithms and compare their energy and speed with GPUs and TPUs.
More about the research
Both projects build probabilistic computers, or p-computers. These machines use probabilistic bits (p-bits), which flip randomly between 0 and 1 in a controllable way. Many p-bits are made from naturally noisy magnetic nanodevices.
Why it matters
Training AI models and checking that they are safe both consume large amounts of energy and computing time. Probabilistic computers built from magnetic devices could run these tasks far more efficiently than today's processors. Such machines could make AI cheaper to use and could help verify neural networks in self-driving cars, aircraft and phones. Built on MRAM technology, they may eventually scale to a million p-bits.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Jifa Tian
University of Wyoming, WY·Physics
Determine the roles and behavior of unconventional magnetic domains, regions where atomic magnetic moments align, in few-layer two-dimensional van der Waals magnets using magnetotransport and low-temperature magnetic force microscopy.
Deliang Fan
Arizona State University Campus Immersion, AZ·Engineering·Early-career
Design, fabricate and validate a hybrid in-memory computing chip that combines RRAM, a non-volatile resistive memory, with SRAM to run deep neural networks with high energy efficiency.
Shantanu Chakrabartty
Washington University in St Louis, MO·Engineering
Model single-electron transport across a Fowler-Nordheim quantum tunneling barrier as the equivalent of a neural spike in NeuroSA, a spiking architecture where neuron firing thresholds follow tunneling physics.
The school
University of California-Santa Barbara: 17 labs on PI Pages, $303M a year on research, 403 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.