Associate Professor Kohei Nakajima (The University of Tokyo), a member of this research project devised a new learning algorithm suitable for physical deep learning by extending a biologically-inspired training algorithm in collaboration with Assistant Professor Inoue (The University of Tokyo) and Nippon Telegraph and Telephone Corporation, and they confirmed its effectiveness.
By applying this learning algorithm to a physical neural network using photonic components, which is expected to be a high-speed machine learning device, they demonstrated that it is possible to carry out efficient computation including the learning process, and achieved the world’s highest performance. It is expected to significantly reduce power consumption and computation time in computing for artificial intelligence in the future.
The research results were published in the online edition of the scientific journal Nature Communications on December 26, 2022.

Bibliography:
[1] Nakajima, Mitsumasa, Katsuma Inoue, Kenji Tanaka, Yasuo Kuniyoshi, Toshikazu Hashimoto, and Kohei Nakajima. 2022. “Physical Deep Learning with Biologically Inspired Training Method: Gradient-Free Approach for Physical Hardware.” Nature Communications 13 (1): 7847. https://doi.org/10.1038/s41467-022-35216-2.

Follow me!