114學年材料系-第七週 專題研討課程 演講公告 (114年10月20日)
2025.10.19
114學年上學期 第七週 專題研討課程 演講公告 (114年10月20日)
題目:
Bridging Devices, Circuits, and Algorithms for Future Neuromorphic Intelligence
演講者: 王超鴻
現職: 國立成功大學敏求智慧運算學院 助理教授
時間: 114年10月20日(一) 15:20~17:10
地點: 成功大學成功校區 三系館 A1307
內容摘要:
Neuromorphic computing, inspired by the efficiency of biological neural systems, is opening new horizons for energy-efficient artificial intelligence and sensing. In this talk, I will highlight recent advances spanning device innovations, circuit design, and algorithm development. On the device front, resistive random-access memory (ReRAM) continues to advance as a candidate for analog synaptic weights, where material engineering and calibration techniques enhance both reliability and efficiency. In addition, Ag-based conductive-bridge RAM (CBRAM) with ultra-low operation voltage characteristics provides an attractive pathway toward low-power synaptic devices. At the circuit level, capacitor-based neuronal analog compute-in-memory (CIM) circuits enable biologically inspired signal integration and efficient neuronal operations. Complementing these device and circuit technologies, piezotronic devices based on novel piezoelectric materials extend neuromorphic hardware into energy harvesting, sensing, and capacitive piezotronic sensor applications. From the algorithmic perspective, we investigate hardware-aware spiking neural networks (SNNs), including brain-like recurrent architectures that exploit neuronal time constants for improved continual learning and network adaptability. Together, these developments chart a course toward future sensory-memory systems that combine optimized devices, innovative circuits, and brain-inspired SNN algorithms to achieve highly efficient and adaptive artificial intelligence.
歡迎各位撥冗參加!
題目:
Bridging Devices, Circuits, and Algorithms for Future Neuromorphic Intelligence
演講者: 王超鴻
現職: 國立成功大學敏求智慧運算學院 助理教授
時間: 114年10月20日(一) 15:20~17:10
地點: 成功大學成功校區 三系館 A1307
內容摘要:
Neuromorphic computing, inspired by the efficiency of biological neural systems, is opening new horizons for energy-efficient artificial intelligence and sensing. In this talk, I will highlight recent advances spanning device innovations, circuit design, and algorithm development. On the device front, resistive random-access memory (ReRAM) continues to advance as a candidate for analog synaptic weights, where material engineering and calibration techniques enhance both reliability and efficiency. In addition, Ag-based conductive-bridge RAM (CBRAM) with ultra-low operation voltage characteristics provides an attractive pathway toward low-power synaptic devices. At the circuit level, capacitor-based neuronal analog compute-in-memory (CIM) circuits enable biologically inspired signal integration and efficient neuronal operations. Complementing these device and circuit technologies, piezotronic devices based on novel piezoelectric materials extend neuromorphic hardware into energy harvesting, sensing, and capacitive piezotronic sensor applications. From the algorithmic perspective, we investigate hardware-aware spiking neural networks (SNNs), including brain-like recurrent architectures that exploit neuronal time constants for improved continual learning and network adaptability. Together, these developments chart a course toward future sensory-memory systems that combine optimized devices, innovative circuits, and brain-inspired SNN algorithms to achieve highly efficient and adaptive artificial intelligence.
歡迎各位撥冗參加!