Harnessing the Power of Memristors for Developing an Energy-Efficient Bayesian Machine

A Bayesian Machine based on Memristors

In the last few decades, machine learning models have improved dramatically in their performance on real-world tasks. However, training and implementing these models still require vast amounts of computational power and energy.

Engineers around the world have therefore been working to develop alternative hardware that could run artificial intelligence models better, in order to promote their use and increase sustainability. These solutions include memristors – memory devices which can store data without using energy.

Researchers at Universite Paris-Saclay-CNRS, Universite Grenoble-Alpes-CEA-LETI, HawAI.tech, Sorbonne Universite, and Aix-Marseille Universite-CNRS have recently created a so-called Bayesian machine (i.e., an AI approach that performs computations based on Bayes’ theorem), using memristors. The proposed system was published in Nature Electronics and found to be much more energy efficient than the hardware currently used.

Source:
https://techxplore.com/news/2023-01-bayesian-machine-based-memristors.html