IAP Seminar(Designing of memristor-based hardware neuromorphic system)
Date : August 21, 2024 11:00 ~ 12:00
Speaker : Prof. Jingon Jang(School of Computer and Information Engineering, Kwangwoon University)
Professor : Prof. Takhee Lee
Location : 56-521
Designing of memristor-based hardware neuromorphic system
Abstract: Recently, the interest in neuromorphic computing has been intensively developed as grown importance of AI application using big-data processing, which could overcome von Neumann bottleneck issues such as the memory wall and immense energy consumption in conventional CMOS computing architecture. In particular, a memristor device capable of crossbar array structure, nanoscale, and non-volatile analog conductance switching states can actualize the ultimate area/energy savings in massively parallel computation such as vector-matrix multiplication (VMM), which is essential for neuromorphic computing in manner of hardware implementation. Here, the several achievements of memristor-based neuromorphic computing system are introduced especially in terms of synaptic plasticity, that is, nonvolatile and analog conductance switching to implement the emulation of neural network perceptron in hardware level [1]. Also, the methodology of real-time data processing for human-machine interfaces are suggested as physical motion recognition system as human-machine interfaces [2], and the application of memristor-based VMM in time-series signal processing is provided to improve the human convenience in traffic system [3].
[1] J. Jang et al., “A learning-rate modulable and reliable TiOx memristor array for robust, fast, and accurate neuromorphic computing”, Advanced Science 9, 2201117 (2022).
[2] H. Cho et al., “Real-time finger motion recognition in free space via skin-compatible electronics”, Nature Electronics 6, 619 (2023).
[3] J. Jang et al., “Active traffic signal decisions using vector-matrix multiplication”, Advanced Intelligent Systems 5, 2200228 (2023).
Abstract: Recently, the interest in neuromorphic computing has been intensively developed as grown importance of AI application using big-data processing, which could overcome von Neumann bottleneck issues such as the memory wall and immense energy consumption in conventional CMOS computing architecture. In particular, a memristor device capable of crossbar array structure, nanoscale, and non-volatile analog conductance switching states can actualize the ultimate area/energy savings in massively parallel computation such as vector-matrix multiplication (VMM), which is essential for neuromorphic computing in manner of hardware implementation. Here, the several achievements of memristor-based neuromorphic computing system are introduced especially in terms of synaptic plasticity, that is, nonvolatile and analog conductance switching to implement the emulation of neural network perceptron in hardware level [1]. Also, the methodology of real-time data processing for human-machine interfaces are suggested as physical motion recognition system as human-machine interfaces [2], and the application of memristor-based VMM in time-series signal processing is provided to improve the human convenience in traffic system [3].
[1] J. Jang et al., “A learning-rate modulable and reliable TiOx memristor array for robust, fast, and accurate neuromorphic computing”, Advanced Science 9, 2201117 (2022).
[2] H. Cho et al., “Real-time finger motion recognition in free space via skin-compatible electronics”, Nature Electronics 6, 619 (2023).
[3] J. Jang et al., “Active traffic signal decisions using vector-matrix multiplication”, Advanced Intelligent Systems 5, 2200228 (2023).