论文信息 - A Scalable Analog Implementation of Multilayer Boltzmann Machines

A Scalable Analog Implementation of Multilayer Boltzmann Machines

A new analog implementation of multilayer Boltzmann machines is described. Its intrinsic scalability has been achieved by original design of low-distorsion analog cells, cascadable random generators and digital interconnectios. Feasibility is demonstrated by the realization of a 1.2 GCPS unit.

Patrick Garda | Eric Belhaire | Hubert Pujol

[1] Stuart Haber,et al. A VLSI-efficient technique for generating multiple uncorrelated noise sources and its application to stochastic neural networks , 1991 .

[2] P. Garda,et al. Design of a linear 'sum-of-product' circuit for Boltzmann machines , 1991, 1991., IEEE International Sympoisum on Circuits and Systems.

[3] Howard C. Card,et al. Cellular automata-based pseudorandom number generators for built-in self-test , 1989, IEEE Trans. Comput. Aided Des. Integr. Circuits Syst..

[4] Robert B. Allen,et al. Relaxation Networks for Large Supervised Learning Problems , 1990, NIPS.

[5] P. Garda,et al. Two Analog Counters for Neural Networks Implementation , 1990, ESSCIRC '90: Sixteenth European Solid-State Circuits Conference.

[6] Koichiro Mashiko,et al. A 336-neuron, 28 K-synapse, self-learning neural network chip with branch-neuron-unit architecture , 1991 .

[7] S. Wolfram. Random sequence generation by cellular automata , 1986 .