Summary
The design of components for a programmable analog neural computer and simulator is described. The machine can be scaled to any size and is composed of three types of interconnected modules, each containing on a VLSI chip arrays of Neurons, modifiable Synapses and Routing Switches. It runs entirely in analog mode but the connection architecture, synaptic gains and time constants as well as neuron parameters are set digitally from a digital host computer. Each neuron has a limited number of inputs and can be connected to any but not all other neurons.
The neuron circuit consists of a rectified summing amplifier, comparator and output driver. Inputs to the neurons are currents, outputs are analog voltages. The following neuron parameters can be adjusted through digital control: threshold (bias), minimum output at threshold and linearity of the transfer function. For the computation of synaptic weights by the host computer on the basis of learning algorithms, time segments of the neuron outputs are multiplexed, converted to digital form and stored in memory.
The synaptic weights are implemented by current mirrors that scale the neuron outputs after they have been converted linearly from a voltage to a current. The weights are set by serial input from the host computer and are stored at each synapse. Dynamic range of the weights extends from 0 to 10 with 5 bit logarithmic resolution; a sixth bit determines the sign. Synaptic time constants are programmed at the inputs to the synapse line.
The routing switches connect vertical and horizontal lines of a cross point array and also can cut these lines. Each switch cell is implemented as a transmission gate connected to one bit of memory. The switches are set by serial input from the host computer.
The machine is intended for real-world, real-time computations such as vision, acoustics or robotics and the development of special purpose neural nets, Even at moderate size of 103 to 105 neurons the computational speed is expected to exceed by orders of magnitude that of any current digital computer.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
References
Mead, C.A. Analog VLSI and Neural Systems. Addison-Wesley, 1989
Raffel, J.I., Mann, J.R., Berger, R., Soares, A.M., Gilbert, S. A Generic Architecture for Wafer-Scale Neuromorphic Systems. IEEE First International Conference on Neural Networks, III-501, San Diego, CA 1987.
Alspector, J., Allen, R.B. A Neuromorphic VLSI Learning System. Advanced Research in VLSI. Proceedings of the 1987 Stanford Conference, 1987.
Tsividis, Y., Satyanarayana, S. Analogue Circuits For Variable-Synapse Electronic Neural Networks. Electronics Letters 23:1313–1314, 1987
Schwartz, D., Howard, R., Hubbard, W. A Programmable Analog Neural Network Chip, IEEE J. of Solid State Circuits, (to be published).
Graf, H.P. et al. VLSI Implementation of a Neural Network Memory with Several Hundreds of Neurons, Neural Networks for Computing, AIP Conference Proceedings, 151:182–187, 1986.
Mueller, P., Van der Spiegel, J., Blackman, D., Chiu, T., Clare, T., Dao, J., Donham C., Hsieh, T., Loinaz, M. Programmable Analog Neural Computer and Simulator, In Advances in Neural Information Processing Systems, Proceedings of the IEEE Conference Denver 1989, D. Touretzky Ed., Morgan Kaufmann, San Mateo, CA.
Chiang, A., CCD Retina and Neural Network Processor, Hardware Implementation of Neural Nets and Synapses, Report on a Workshop. P. Mueller ed. San Diego 1988.
Shoemaker, PA., Shimabukuro, R. A Modifiable Weight Circuit for Use in Adaptive Neuromorphic Networks. Neural Networks, 1,Suppl.I, 409, 1988.
Sage, J.P., Thompson, K., Withers, R.S., An artificial Neural Network Integrated Circuit Based on MNOS/ CCD Principles, Proc. Conf. Neural Networks for Computing, AIP 151, 381 1986.
Rudnick, M., Hammerstrom, D., An Interconnect Structure for Waferscale Neurocomputers, Neural Networks, 1,Suppl.I, 404, 1988.
White, J., Furman, B., Abidi, AA., Baker, R.L., Mathur, B., Wang, H.T., Parallel Architecture for 2D Gaussian Convolution of Images, Neural Networks, 1, Suppl. I, 415, 1988
Furman, B., White, J., Abidi, AA., CMOS Analog IC Implementing the Back Propagation Algorithm, Neural Networks, 1,Suppl.I, 381, 1988.
Akers, LA., Walker, M.R., Training a Limited Interconnect Feedforward Neural Array, Neural Networks, 1, suppl.I 1381, 1988.
Mead, C. and Mahowald, MA. A Silicon Model of Early Visual Processing. Neural Networks, 1:91–97, 1988.
Lyon, R.F. and Mead, C.A. An Analog Electronic Cochlea. IEEE Trans. Acoust., Speech, Signal Processing, 36:1119–1134, 1988.
Lazarro, J.P., Ryckebusch, S., Mahowald, M.A., and Mead, C.A., Winner Take All Networks of O(n) Complexity, Proc.JEEE Conference on Neural Information Processing Systems, Denver, CO, 1988.
Mueller, P., Lazzaro, J. A Machine for Neural Computation of Acoustical Patterns. Neural Networks for Computing, AIP Conference Proceedings, 151:321–326, 1986.
Mueller, P., Martin, T., Putzrath, F., General Principles of Operation in Neuron Nets with Application to Acoustical Pattern Recognition. Biological Prototypes and Synthetic Systems, E.E. Bernard and M.R. Kare Eds., Plenum Press New York. 1962.
Rumelhart, D.E., Hinton, G.E., Williams, RJ., Learning Internal Representations by Error Propagation, Parallel Distributed Processing V. 1, 319 MIT Press 1986.
Bult, K. and Wallinga, H. A Class of Analog CMOS Circuits Based on the Square Law Characteristics of an MOS Transistor in Saturation. IEEE J. Solid-State Circuits, SC-22:3657, 1987.
Mueller P., Principles of Temporal Pattern Recognition in Artificial Neuron Nets, Artificial Intelligence, S 142, The Institute of Electrical and Electronic Engineers, Inc., New York 1963.
Mueller, P., Computation of Temporal Pattern Primitives in a Neural Net for Speech Recognition, In. Report on a Workshop “Hardware Implementation of Neuron Nets and Synapses”, P. Mueller Ed., San Diego CA., 1988..
Mueller, P. Computation of Temporal Pattern Primitives in a Neural Net for Speech Recognition. Abstracts of the First Annual INNS Meeting, Boston, MA, 1(1):308, 1988.
Hu, V., Kramer, A., Ko, P.K., EEPROMS as Analog Storage Devices for Neural Nets, Neural Networks, 1 Suppl. I 1385, 1988.
Thakoor, A., Modifiable Synapses, In Report on a workshop “Hardware Implementation of Neuron Nets and Synapses”, P. Mueller Ed., San Diego CA., 1988..
Mueller, P., Blackman, D., Spiro, P., Furman, R.E., Neural computation of visual images, IEEE First Annual Conference on Neural Networks, 4, 75, San Diego CA, 1987.
Mueller, P., Blackman, D., Furman, R.E., Neural computation of Visual Images, In: An Introduction to Neural and Electronic Networks, S.F. Zornetzer, ed., Academic Press, 1989.
Mueller, P., Neural computation of Pattern Primitives, AIP Conference on Neuron Networks Snowbird Utah 1988.
Sivilotti, M., A Dynamically Configurable Architecture for Prototyping Analog Circuits, Advanced Research in VLSI, Proc. of the Fifth MIT Conference. Allen and Leighton eds., MIT Press, Cambridge MA. 1988.
Press, W.H., Flannery, B.P., Teukolsky, S.A., Vetterling, W.T., Numerical Recipes, Chapter 15,Cambridge University Press, 1986.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1989 Kluwer Academic Publishers
About this chapter
Cite this chapter
Mueller, P. et al. (1989). Design and Fabrication of VLSI Components for a General Purpose Analog Neural Computer. In: Mead, C., Ismail, M. (eds) Analog VLSI Implementation of Neural Systems. The Kluwer International Series in Engineering and Computer Science, vol 80. Springer, Boston, MA. https://doi.org/10.1007/978-1-4613-1639-8_6
Download citation
DOI: https://doi.org/10.1007/978-1-4613-1639-8_6
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4612-8905-0
Online ISBN: 978-1-4613-1639-8
eBook Packages: Springer Book Archive