Analog VLSI neural network implementations of hardware annealing and winner-take-all functions

Hardware annealing and winner-take-all (WTA) functions have been implemented in 2.0- mu m technology. The hardware annealing technique has been demonstrated using a 4*4 synapse network. Measurement results of a new WTA circuit are presented. The WTA circuit uses transistors biased in saturation to achieve high-speed performance. Since the comparison among the inputs is performed on one common signal line, the circuit can be easily extended to a larger dimension with that common signal line connected throughout the entire circuit. The new high-speed analog winner-take-all circuit can be extended linearly to at least 1024 inputs.<<ETX>>

[1]  J. J. Paulos,et al.  Artificial neural networks using MOS analog multipliers , 1990 .

[2]  Bing J. Sheu,et al.  A trainable analog neural chip for image compression , 1991, Proceedings of the IEEE 1991 Custom Integrated Circuits Conference.

[3]  Bing J. Sheu,et al.  A compact and general-purpose neural chip with electrically programmable synapses , 1990, IEEE Proceedings of the Custom Integrated Circuits Conference.

[4]  Tobi Delbrück,et al.  Cooperative Stereo Matching Using Static and Dynamic Image Features , 1989, Analog VLSI Implementation of Neural Systems.

[5]  John J. Hopfield,et al.  Simple 'neural' optimization networks: An A/D converter, signal decision circuit, and a linear programming circuit , 1986 .

[6]  Mohammed Ismail,et al.  An analog MOS implementation of the synaptic weights for feedback neural nets , 1989, IEEE International Symposium on Circuits and Systems,.

[7]  C. Tomovich,et al.  MOSIS - A gateway to silicon , 1988, IEEE Circuits and Devices Magazine.

[8]  Bing J. Sheu,et al.  Hardware annealing in analog VLSI neurocomputing , 1990 .