VLSI Bit-Serial Neural Networks
A synthetic neural network can be viewed as a large parallel array of n 2 synaptic operators (for n neurons) that is able to model some of the brain’s characteristics. The VLSI neural network described, functions with bit-serial, two’s complement arithmetic and uses a single phase clocking technique operating at a minimum of 20 MHz (McGregor et al 1987).
KeywordsSynaptic Weight Hardware Accelerator Neural State Control Circuitry Clock Scheme
Unable to display preview. Download preview PDF.
- Hopfield, J. J., “Neural Networks with Emergent Collective Computational Abilities,” Proceedings of the National Academy of Science, USA, vol. 79, pp. 2554–2558, 1982.Google Scholar
- McGregor, M.S., Denyer, P.B. and Murray, A.F., “A Single — Phase Clocking Scheme for CMOS VLSI,” Advanced Research in VLSI: Proceedings of the 1987 Stanford Conference, 1987.Google Scholar
- Murray, A.F., Smith, A.V.W. and Butler, Z. F., “Bit-serial Neural Networks,” IEEE Conf. on Neural Infomation Processing Systems — Natural and Synthetic, Denver, 1987.Google Scholar
- Rumelhart, D.E., Hinton, G.E. and Williams, R.J., “Learning Internal Representations by Error Propagations”, Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 1, pp. 318–362, 1986.Google Scholar