Skip to content

A fast and energy-efficient sparse icing machine for solving computationally difficult problems

A fast and energy-efficient sparse icing machine for solving computationally difficult problems

tech innovation 2022

The team implemented a 5000 p-bit probabilistic computer on state-of-the-art field programmable gate arrays. Credits: Adit et al

In recent years, engineers have been trying to develop new computers and tools that can help solve challenging real-world problems faster and more efficiently. Some of the most promising of these are icing machines (IMs), physics-based systems designed to tackle complex optimization problems.

Researchers from the University of California and the University of Messina have recently developed a rare icing machine architecture that can work on classical and existing computer hardware. This architecture, presented in a paper published in Prakriti ElectronicsIt was found to be significantly faster than standard optimization methods running on a central processing unit.

“Building domain-specific, quantum-inspired architectures has become an important area of ​​research with the slowing of Moore’s Law,” Kerem Camasari, one of the researchers who conducted the study, told TechExplore. “The primary purpose of this work was to extend our earlier work on probabilities, or p-bits, conceptually between bits and qubits.”

In 2019, Camasari and colleagues showed that eight p-bit networks based on nano devices can help solve some difficult optimization problems in energy-efficient ways. In their new paper, they expanded their network to include 5,000 p-bits using classical CMOS technology. It is a pioneering technology used to manufacture integrated circuit (IC) chips and other electronic components.

The team found that increasing the p-bits of their architecture resulted in higher speeds and performance, allowing it to tackle more complex optimization problems more efficiently. Furthermore, his architecture was found to outperform the state-of-the-art, classical approaches that have been widely used for decades.

“What is particularly promising about our recent work is that the architecture we developed here can be applied to the spintronics technology,” Giovanni Finocchio, another researcher involved in the study, told TechExplore. “As we have shown before yearP-computing can be highly spintronics compatible and orders of magnitude further improvements in speed and scalability can be achieved in integrated magnetic p-computers.”

The main idea enabling parallelization was to convert optimization problems into less dense (sparsified) networks at the cost of additional p-bits. Credits: Adit et al

The sparse ising machine developed by Camsari, Finocchio and their colleagues is based on the idea that when making probabilistic decisions, equality comes rarely. In other words, his approach assumes that consulting less reliable sources allows us to make an informed decision faster and more efficiently than consulting multiple parties.

“We have invented techniques that can take any difficult optimization problem and turn it into a sparse network for parallel sampling,” said Navid Anjum Adit, a researcher involved in the study. “A unique feature of our architecture is its performance (probabilistic updates per second) that scales linearly with the number of p-bits in the system, is extremely unusual, and is the highest level of parallelism we can hope to achieve.” Huh.”

The findings gathered by this team of researchers highlight the potential of sparse icing machines, even when running on traditional computer hardware. In fact, he found that his icing machine could tackle optimization problems as well, if not better, than many cutting-edge classical techniques when running on existing p-computers.

Andrea Grimaldi, one of the researchers who conducted the study, told TechExplore, “A particularly impressive example was solving the integer factorization problem for very large numbers (up to 32-bits), which would be useful for anyone attempting this problem.” was much larger than other probabilistic solvers.” TechXplore. “However, we should mention that there are many non-probabilistic algorithms in factorization and these may be more efficient than our approach. Our aim was to see how our machine could solve extremely difficult optimization problems, giving us other probabilities solver, classical or quantum.”

In the future, the sparse ising machine architecture developed by Camsari, Finocchio, Aadit, Grimaldi and their colleagues could be applied to many other real-world optimization problems. In their next studies, the researchers plan to increase their pi-computer from the 5,000 p-bits to the 50,000-100,000 p-bits they are currently evaluating.

“We are deeply interested in designing new algorithms and architectures, but also harnessing the power and promise of emerging technology such as magnetic nanodevices,” Camassari said. “We are constantly looking for new applications of p-computers in quantum computing as well as artificial intelligence.”

P-Computer Capability

more information:
Anjum Adit et al, Massively Parallel Probabilistic Computing with Sparse Ising Machines, Prakriti Electronics (2022). DOI: 10.1038/s41928-022-00774-2

© 2022 Science X Network

Citation: a fast and energy-efficient sparse icing machine to solve computationally difficult problems (2022, 23 June) Received 23 June 2022

This document is subject to copyright. No part may be reproduced without written permission, except for any fair use for the purpose of personal study or research. The content is provided for information purposes only.

credit source

A fast and energy-efficient sparse icing machine for solving computationally difficult problems

#fast #energyefficient #sparse #icing #machine #solving #computationally #difficult #problems

if you want to read this article from the original credit source of the article then you can read from here

Shopping Store 70% Discount Offer

Leave a Reply

Your email address will not be published.