HOME > Research > New computing and Neuromorphic Research
New computing and Neuromorphic Research
- Research on the storage class memory for fast data processing speed that can prevent data bottleneck has been conducted vigorously. Among the candidates, ReRAM (resistive switching random access memory) and PcRAM (phase-change random access memory) have been spotlighted due to their ultra-high density integration and low power consumption. Our group has expanded the research area to the application study of single ReRAM or PcRAM devices and memristor crossbar array based on research results and experiences of the switching mechanism and a memory structure capable of high-density integration study that has been conducted so far.
- Neuromorphic computing that mimics the human neural systems has been developed to solve the problems of traditional computing systems especially regarding computing speed and power consumption. Our group has studied hardware implementation using nonvolatile next-generation memories. Artificial synapse transmitting the external stimuli to the internal nerve system is the most representative device being studied in our lab. Also, time kernels and convolutional kernels are under research for convolutional neural networks (CNN) development. Time-varying data processing based on the resistive switching device was suggested and published in Nature Communications (2021).
- The emerging technology called in-memory-computing has received attention since data migration between memory and computing unit is unnecessary, enhancing data processing speed and eventually solving the bottleneck issue. Moreover, owing to its gradual conductance change, digital and analog logic-in-memory computing can be realized with ReRAM. Our group conducts research on logic device tuning with line resistance and local electrode engineering, designing system architecture, and improving reliability to enhance the logic in-memory computation efficiency.
- Our group also researches security applications exploiting the volatility of the Cu-Te alloy-based conductive bridge random access memory (CBRAM). The memristor-based true random number generator (TRNG) experimentally proved the feasibility of adopting the memristor in security applications for the IoT era. Recently, we are expanding the research area to a new method of probabilistic computing. The memristor-based probabilistic-bit (p-bit) computing can perform both forward and invertible operations, showing the possibility of expanding its uses for complex problems such as integer factorization.