«Выносливость» устройств для нейроморфных компьютеров повысили учёные МГУ и Курчатовского института

Experts increased the number of switching cycles of memristors by 1.5 times — a step towards more reliable computing

Moscow scientists from Moscow State University named after M. V. Lomonosov and NRC “Kurchatov Institute” have increased the efficiency of memristors — key elements for neuromorphic computers that mimic the operation of biological neural networks. According to the press service of the university, the researchers managed to increase the number of switching cycles of such resistors by 1.5 times.

Memristors differ from conventional resistors in that their resistance depends on previous current loads. This gives the devices “memory” and allows them to change the states recorded in them, bringing their functions closer to nerve endings.

A group from the Faculty of Physics of Moscow State University and the D. V. Skobeltsyn Institute of Nuclear Physics, together with colleagues from the Kurchatov Institute, improved the characteristics of memristors by irradiation with α‑particles.

According to the researchers, controlled formation of defects under radiation exposure led to a significant improvement: the number of stable resistive states increased almost threefold, the contrast between high and low resistance increased by more than two times, and the endurance of the devices — the number of switching cycles — increased by 1.5 times.

Read also on the topic:

Breakthrough in oil exploration occurred, thanks to the Moscow State University supercomputer

More than 2,500 Russian scientists use the power of Moscow State University supercomputers

A new quantum encryption protocol has been developed at Moscow State University

Sources
TASS

Now on home