论文信息 - Precision Requirements for Analog Weight Adaptation Circuitry for Single-Layer Nets

Precision Requirements for Analog Weight Adaptation Circuitry for Single-Layer Nets

When implementing neural networks with learning capability in analog hardware, parasitic weight adaptation due to offsets, leakage and charge injection may occur. These effects can be modeled as an extra constant weight adaptation component for each adaptation cycle. An analysis is given of the effect of constant weight adaptation on the learning behavior in sections 6.3 and 6.4. A worst case estimation of the maximum tolerable constant weight adaptation is given in section 6.5.

Anne-Johan Annema

[1] Daniele D. Caviglia,et al. An Experimental Analog VLSI Neural Chip with On-Chip Back-Propagation Learning , 1992 .

[2] Geoffrey E. Hinton,et al. Learning internal representations by error propagation , 1986 .

[3] Daniele D. Caviglia,et al. Effects of weight discretization on the back propagation learning method: algorithm design and hardware realization , 1990, 1990 IJCNN International Joint Conference on Neural Networks.

[4] Yiwen Wang. A modular analog CMOS LSI for feedforward neural networks with on-chip BEP learning , 1993, 1993 IEEE International Symposium on Circuits and Systems.

[5] James L. McClelland,et al. Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations , 1986 .