Reliable Design of Neuromorphic Computers

The application field of machine learning, represented by deep learning, is rapidly expanding. Since some applications require a great deal of time and power consumption, neuromorphic computers, which are hardware for machine learning, are expected to be used, and research on new devices and new architectures for them is actively conducted.

Research on neuromorphic computers using new devices, such as variable resistance memory ReRAM, ferroelectric memory FeRAM, magnetoresistive memory MRAM, and other nonvolatile memories, has been active. The Dependable Systems Laboratory is working on the study of the dependability of neuromorphic computers using ReRAM, which utilizes a passive device with variable resistance called a memristor.