论文信息 - Design of CMOS quadratic neural networks

Design of CMOS quadratic neural networks

In this work, we introduce an analog implementation technique for a class of quadratic artificial neural networks, in which the intrinsic quadratic characteristic of an MOS transistor in its saturation region has been employed. Test chips implemented in a 1.2 /spl mu/m CMOS technology have shown full functionality as designed and demonstrate circular discriminating surfaces.

[1] Robert M. Farber,et al. How Neural Nets Work , 1987, NIPS.

[2] E. Seevinck,et al. A versatile CMOS linear transconductor/square-law function , 1987 .

[3] Dennis J. Volper,et al. Quadratic function nodes: Use, structure and training , 1990, Neural Networks.

[4] Ashok Nedungadi,et al. Design of linear CMOS transconductance elements , 1984 .

[5] Donald F. Specht,et al. Probabilistic neural networks , 1990, Neural Networks.