Presentation Information

[10a-N405-3]Demonstration of Tlerance to Parameter Variations in Superconducting ReLU Neuron Circuit

〇Yuto Ueno1, Yuki Hironaka2, Nobuyuki Yoshikawa1,2, Yuki Yamanashi1,2 (1.Yokohama National Univ., 2.Yokohama National Univ. IAS)

Keywords:

SFQ circuit,Neuromorphic Computing,ReLU

Attempts have been made to improve power consumption and processing speed compared to conventional software-based implementations by creating circuits specifically for artificial neural networks. However, neuron circuits designed in previous research are not suitable for large-scale networks due to variations in output characteristics caused by manufacturing errors in the circuits. This research demonstrated a circuit that is resilient to parameter fluctuations due to manufacturing errors by measuring it across multiple chips.