On-Chip Compensation of Device-Mismatch Effects in Analog VLSI Neural Networks

Device mismatch in VLSI degrades the accuracy of analog arithmetic circuits and lowers the learning performance of large-scale neural networks implemented in this technology. We show compact, low-power on-chip calibration techniques that compensate for device mismatch. Our techniques enable large-scale analog VLSI neural networks with learning performance on the order of 10 bits. We demonstrate our techniques on a 64-synapse linear perceptron learning with the Least-Mean-Squares (LMS) algorithm, and fabricated in a 0.35

• Type: Conference
• Year: 2004
• Where: NIPS
• Authors: Miguel Figueroa, Seth Bridges, Chris Diorio
• Keywords: Device Mismatch, Large-scale Analog Vlsi, Large-scale Neural Networks, NIPS 2004, NIPS 2007