Newsgroups: comp.ai.neural-nets From: srctran@world.std.com (Gregory Aharonian) Subject: Neural network patent abstracts posting Message-ID: Organization: The World Public Access UNIX, Brookline, MA Date: Tue, 12 Jan 1993 21:07:41 GMT Lines: 391 The following is my periodic posting of the abstracts and stat data for new neural network patents. The full text and diagrams of any of these patents can be ordered from the patent office for $3 in printed form. Also, for $20, I will provide a machine readable version of the abstract, text, claims and references for any neural network patent. Greg Aharonian Source Translation & Optimization srctran@world.std.com 617-489-3727 ============================================================================= For a very good article on the legal issues involving neural network patents, see the article "Intellectual Property Protection for Neural Networks" by Donald Wenskay, Neural Networks, 3, pp. 229-236, 1990. ============================================================================= For those interested in applying for a neural network patent, I freely provide a LaTeX template for a patent application. Please email a request, and I'll send it to you. Contains most of the boilerplate you need. ============================================================================= To date, I have yet to see any articles in the trade journals on either companies signing cross-licensing agreements to share their neural network patents, or patent infringement lawsuits dealing with neural network patents. To me, this means that no one is making much money with neural networks (other than selling software), and that the financial return on acquiring a patent is less than the cost of applying for and maintaining a patent. If anyone hears of either occurrence, please let me know. ============================================================================= 5,168,352 [IMAGE AVAILABLE] Dec. 1, 1992 Coloring device for performing adaptive coloring of a monochromatic image INVENTOR: Motohiko Naka, Kawasaki, Japan Mie Saitoh, Kawasaki, Japan Takehisa Tanaka, Tokyo, Japan Kunio Yoshida, Kawasaki, Japan ASSIGNEE: Matsushita Electric Industrial Co., Ltd., Osaka, Japan APPL-NO: 07/480,456 DATE FILED: Feb. 15, 1990 FRN-PRIOR: Japan 1-36779 Feb. 16, 1989 INT-CL: [5] H04N 9/02 US-CL-ISSUED: 358/81, 82, 75; 382/15 US-CL-CURRENT: 358/81, 75, 82; 382/15 SEARCH-FLD: 358/81, 82, 75; 364/513; 382/14, 15 REF-CITED: U.S. PATENT DOCUMENTS 4,760,604 7/1988 Cooper 364/715.01 4,926,250 5/1990 Konishi 358/81 OTHER PUBLICATIONS "A Wafer Scale Integration Neural Network Utilizing Completely Digital Circuits" by Moritoshi Yasunaga et al., IEEE International Joint Conference on Neural Networks, Washington, D.C., Jun. 1989. Richard P. Lippmann, An Introduction to Computing with Neural Nets, IEEE ASSP Magazine Apr. 1987. ART-UNIT: 262 PRIM-EXMR: James J. Groody ASST-EXMR: Sherrie Hsia LEGAL-REP: Lowe, Price, LeBlanc & Becker ABSTRACT: A coloring device includes an image sampling device for sampling an input signal block representing a group of n.times.m pixels of a monochromatic image and for outputting first signals representing the sampled pixels of the input signal block of the monochromatic image; and artificial neural network, a connection for providing to the artificial neural network, substantially simultaneously, pattern information on patterns to be contained in the monochromatic image and color information on first data indicating colors given to the patterns indicated by the pattern information prior to generation of a color image signal, the artificial neural network having internal state parameters which are adaptively optimized by using a learning algorithm prior to the generation of a color image, the artificial neural network operating for receiving data representing the first signal, for determining which of colors preliminarily and respectively assigned to patterns to be contained in the group of pixels of the monochromatic image represented by the input signal block is given to a pattern actually contained in the group of pixels represented by the input signal block and for outputting second signals representing second data on three primary colors which are used to represent the determined colors given to the patterns actually contained in the group of pixels represented by the input signal block; and a color image storing device for receiving the second signals outputted from the artificial neural network, for storing the received second signals in locations thereof corresponding to the positions of the pixels represented by the input signal block and for outputting third signals representing the three primary color component images of the pixels represented by the input signal block; wherein the image sampling device further functions for scanning the whole of the monochromatic image by generating successive input signal blocks representing successive groups n.times.m pixels to be sampled, thereby outputting third signals for all pixels of the monochromatic image. 10 Claims, 4 Drawing Figures EXMPL-CLAIM: 1 NO-PP-DRAWING: 3 ============================================================================== 5,168,550 [IMAGE AVAILABLE] Dec. 1, 1992 Neutral network with plural weight calculation methods and variation of plural learning parameters INVENTOR: Shigeo Sakaue, Takarazuka, Japan Toshiyuki Kohda, Takatsuki, Japan Yasuharu Shimeki, Suita, Japan Hideyuki Takagi, Kyoto, Japan Hayato Togawa, Tokyo, Japan ASSIGNEE: Matsushita Electric Industrial Co., Ltd., Osaka, Japan APPL-NO: 07/481,330 DATE FILED: Feb. 20, 1990 FRN-PRIOR: Japan 1-43730 Feb. 23, 1989 Japan 1-47610 Feb. 28, 1989 INT-CL: [5] G06F 15/18 US-CL-ISSUED: 395/23 US-CL-CURRENT: 395/23 SEARCH-FLD: 364/513; 395/23 REF-CITED: OTHER PUBLICATIONS Improving the Learning Rate of Back-Propagation with the Gradient Reuse Algorithm; Hush et al; IEEE Inter. Conf. on Neutral Networks; Jul. 24-27, 1988; pp. I-441 to I-446. Learning Internal Representations by Error Propagation; Rumelhart et al.; Parallel Distributed Processing, vol. 1, Foundations; MIT Press; 1986; pp. 318-362. "Learning representations by back-propagating errors", David Rumelhart et al.; Nature, vol. 323, Oct. 1986, pp. 533-536. DUTTA, "Bond Ratings: A non-conservative application of neutral networks", IEEE International Conference on Neutral Networks, vol. 2, pp. 443-450, Jul. 24, 1988. Watrous, "Learning algorithms for connectionist networks: applied gradient methods of nonlinear optimization", IEEE First National Conference on Neutral Networks, vol. 2, pp. 619-628, Jun. 21, 1987. JACOBS, "Increased rates of convergence through learning rate adaption", Neutral Networks, vol. 1, No. 4, pp. 295-307, 1988. ART-UNIT: 238 PRIM-EXMR: Allen R. MacDonald LEGAL-REP: Stevens, Davis, Miller & Mosher ABSTRACT: An iterative learning machine uses, as a direction of changing iterative weight, a conjugate gradient direction in place of the conventional steepest descent direction, thereby saving time. Learning rates are set dynamically. Error calculation for plural learning rates, with respect to a certain weight changing direction, are accomplished by storing a product-sum of the input signals and weights in a hidden layer and a product-sum of the input signals and the weight changing direction in the hidden layer. When the learning falls into a non-effective state where further iteration does not effectively reduce an error, the weights are adjusted in order to restart the learning. 31 Claims, 18 Drawing Figures EXMPL-CLAIM: 1 NO-PP-DRAWING: 17 ============================================================================== 5,168,551 [IMAGE AVAILABLE] Dec. 1, 1992 MOS decoder circuit implemented using a neural network architecture INVENTOR: Ho-sun Jeong, Taegu, Republic of Korea ASSIGNEE: Samsung Electronics Co., Ltd., Kyunggi-do, Republic of Korea APPL-NO: 07/573,408 DATE FILED: Aug. 28, 1990 FRN-PRIOR: Republic of Korea 90-4172 Mar. 28, 1990 INT-CL: [5] G06F 15/18 US-CL-ISSUED: 395/27, 24; 364/602, 807 US-CL-CURRENT: 395/27; 364/602, 807; 395/24 SEARCH-FLD: 307/201, 494, 498, 529; 395/24, 27; 364/602, 807 REF-CITED: U.S. PATENT DOCUMENTS 4,876,534 10/1989 Mead et al. 340/825.95 4,904,881 2/1990 Castro 307/201 4,956,564 9/1990 Holler et al. 307/201 4,962,342 10/1990 Mead et al. 307/201 4,978,873 12/1990 Shoemaker 307/498 4,988,891 1/1991 Mashiko 307/201 OTHER PUBLICATIONS McClelland et al., Explorations in Parallel Distributed Processing: A Handbook of Models, Programs, and Exercises, The MIT Press, 1988, pp. 83-99. Walker et al., "A CMOS Neural Network for Pattern Association", IEEE Micro, Oct. 1989, pp. 68-74. Salam et al., "A Feedforward Neural Network for CMOS VSLI Implementation", Midwest Sympos. on Cir. Syst., 1990, pp. 489-492. Graf et al., "VLSI Implementation of a Neural Network Model", Computer, Mar. 1988, pp. 41-49. Tanenbaum, A. S., Structured Computer Organization, Prentice-Hall, Inc., 1984, pp. 121-122. ART-UNIT: 238 PRIM-EXMR: Michael R. Fleming ASST-EXMR: Robert Downs LEGAL-REP: Cushman, Darby & Cushman ABSTRACT: A decoder circuit based on the concept of a neural network architecture has a unique configuration using a connection structure having CMOS inverters, and PMOS and NMOS bias and synapse transistors. The decoder circuit consists of M parallel inverter input circuit corresponding to an M-bit digital signal and forming an input neuron group, a 2.sup.M parallel inverter output circuit corresponding to 2.sup.M decoded outputs and forming an output neuron group, and a synapse group connected between the input neuron group and the output neuron group responsive to a bias group and the M-bit digital original for providing a decoded output signal to one of the 2.sup.M outputs of the output neuron group when a match is detected. Hence, only one of the 2.sup.M outputs will be active at any one time. 6 Claims, 6 Drawing Figures EXMPL-CLAIM: 1 NO-PP-DRAWING: 2 ============================================================================== 5,170,071 [IMAGE AVAILABLE] Dec. 8, 1992 Stochastic artifical neuron with multilayer training capability INVENTOR: Gregory A. Shreve, Redondo Beach, CA ASSIGNEE: TRW Inc., Redondo Beach, CA (U.S. corp.) APPL-NO: 07/716,717 DATE FILED: Jun. 17, 1991 INT-CL: [5] G06F 15/18 US-CL-ISSUED: 307/20; 395/27 US-CL-CURRENT: 307/201; 395/27 SEARCH-FLD: 395/27; 307/201 REF-CITED: U.S. PATENT DOCUMENTS 3,341,823 9/1967 Connelly 307/201 X 3,691,400 9/1972 Askew 307/201 3,950,733 4/1976 Cooper et al. 340/172.5 4,518,866 5/1985 Clymer 307/201 4,591,980 5/1986 Huberman et al. 364/200 4,773,024 9/1988 Faggin et al. 364/513 4,807,168 2/1989 Moopenn et al. 364/602 4,809,193 2/1989 Jourjine 364/513 4,893,255 1/1990 Tomlinson 364/513 4,918,618 4/1990 Tomlinson 364/513 4,989,256 1/1991 Buckley 395/27 X OTHER PUBLICATIONS Tomlinson, Jr., Walker and Sivilotti, "A Digital Neural Network Architecture for VLSI", IJCNN 1990, Jun. 1990, San Diego, Calif. Walker and Tomlinson, Jr., "DNNA: A Digital Neural Network Architecture", INNC, Jul. 9-13, 1990. P. C. Patton, "The Neural Semiconductor NU32/SU3232 Chip Set," the Superperformance Computing Service Brief No. 36, Feb. 1990. R. Colin Johnson, "Digital Neurons Mimic Analog," Electronic Engineering Times, Feb. 12, 1990. Nguyen, Dziem and Holt, Fred, "Stochastic Processing in a Neural Network Application," IEEE First International Conference on Neural Networks, San Diego Calif., Jun. 21-24, 1987, pp. 281-291. Gaines, Brian R., "Uncertainty as a Foundation of Computational Power in Neural Neworks," IEEE First International Conference on Neural Networks, San Diego Calif., Jun. 21-24, 1987, pp. 51-57. Van den Bout, David E. and Miller, T. K., "A Stochastic Architecture for Neural Nets," pp. 481-488. Rumelhart, David E. et al., "Learning Representations by Back-Propagating Errors," Nature, vol. 323, Oct. 9, 1986, pp. 533-536. Rumelhart, David E. et al., "Learning Internal Representations by Error Propagation," Institute for Cognitive Science (ICS) Report 8506, Sep. 1985. ART-UNIT: 259 PRIM-EXMR: David Hudspeth LEGAL-REP: James M. Steinberger, G. Gregory Schivley, Ronald L. Taylor ABSTRACT: A probabilistic or stochastic artificial neuron in which the inputs and synaptic weights are represented as probabilistic or stochastic functions of time, thus providing efficient implementations of the synapses. Stochastic processing removes both the time criticality and the discrete symbol nature of traditional digital processing, while retaining the basic digital processing technology. This provides large gains in relaxed timing design constraints and fault tolerance, while the simplicity of stochastic arithmetic allows for the fabrication of very high densities of neurons. The synaptic weights are individually controlled by a backward error propagation which provides the capability to train multiple layers of neurons in a neural network. 8 Claims, 3 Drawing Figures EXMPL-CLAIM: 1 NO-PP-DRAWING: 3 ============================================================================== 5,170,463 [IMAGE AVAILABLE] Dec. 8, 1992 Neuro-computer INVENTOR: Yoshiji Fujimoto, Nara, Japan Naoyuki Fukuda, Nara, Japan Toshio Akabane, Tenri, Japan ASSIGNEE: Sharp Kabushiki Kaisha, Osaka, Japan APPL-NO: 07/885,239 DATE FILED: May 20, 1992 REL-US-DATA: Continuation of Ser. No. 456,649, Dec. 27, 1989, abandoned. FRN-PRIOR: Japan 63-330971 Dec. 29, 1988 Japan 1-24307 Feb. 1, 1989 Japan 1-127274 May 19, 1989 INT-CL: [5] G06F 15/16 US-CL-ISSUED: 395/11, 24, 27; 364/DIG.2, DIG.1 US-CL-CURRENT: 395/11; 364/DIG.1, DIG.2; 395/24, 27 SEARCH-FLD: 395/11, 24, 27; 364/200, 900 REF-CITED: U.S. PATENT DOCUMENTS 4,514,807 4/1985 Nogi 364/200 4,633,472 12/1986 Krol 364/200 4,644,496 2/1987 Andrews 364/513 4,660,166 4/1987 Hopfield 364/807 4,709,327 11/1987 Hillis et al. 364/513 4,739,476 4/1988 Fiduccia 364/513 4,766,534 8/1988 De Benedictis 364/513 4,796,199 1/1989 Hammerstrom et al. 364/513 4,809,193 2/1989 Jourjine 364/513 4,811,210 3/1989 McAulay 364/513 4,858,147 8/1989 Conwell 364/200 4,891,782 1/1990 Johnson 364/786 4,908,751 3/1990 Smith 364/513 4,912,647 3/1990 Wood 364/513 4,918,617 4/1990 Hammerstrom et al. 364/513 4,918,618 4/1990 Tomlinson, Jr. 364/513 4,920,487 4/1990 Baffes 364/200 4,942,517 7/1990 Cok 364/200 4,951,239 8/1990 Andes et al. 364/807 OTHER PUBLICATIONS The Computer Journal, vol. 30, No. 5, Oct. 1987, pp. 413-419; B. M. Forrest et al: "implementing neural network models on parallel computers". Proceedings of the 1983 International Conference on Parallel Processing, Columbus, Ohio, Aug. 1983, pp. 95-105. IEEE, New York US; T. Hoshino et al: "Highly parallel processor array PAX for wide scientific applications". 1988 Cern School of Computing, Oxford, Aug. 1988, pp. 104-126; P. C. Treleaven: "Parallel architectures for neurocomputers". IEEE Communications Magazine, vol. 26, No. 6, Jum. 1988, pp. 45-50; T. G. Robertazzi: "Toroidal networks". Proceedings of the IEEE 1988 National Aerospace and Electronics Conference NAECON '88, Dayton, May 1988, vol. 4, pp. 1574-1580; B. C. Deer et al: "Parallel processor for the simulation of adaptive networks". D. A. Pomerleau et al, "Neural Network Simulation at Warp Speed: How We Got 17 Million Connections Per Second," Proceedings of the IEEE ICNN, San Diego, Calif., Jul. 1988, vol. II, pp. 143-150. S. Y. Kung and J. N. Hwang, "Parallel Architectures for Artificial Neural Networks," Proceedings of the IEEE ICNN, San Diego, Calif., Jul. 1988, vol. II, pp. 165-172. S. Y. Kung and J. N. Hwang, "Ring Systolic Designs for Artificial Neural Nets," Abstracts of the First Annual INNS Meeting, Boston, Mass., Sep. 1988, p. 390. G. Blelloch and C. R. Rosenberg, "Network Learning on the Connection Machine," Proceedings of the IJCAI, Milano, Italy, Aug. 1987, pp. 323-326. A. Johannet et al, "A Transputer-Based Neurocomputer," Parallel Programming of Transputer Based Machines: Proceedings of the 7th OCCAM User Group Technical Meeting Grenoble, France, Sep. 1987, pp. 120-127. D. Suter and X. Deng, "Neural Net Simulation on Transputers," Australian Transputer and OCCAM User Group Conference Proceedings, Royal Melbourne Institute of Technology, Jun. 1988, pp. 43-47. Jim Baily and Dan Hammerstrom, "Why VLSI Implementation of Associative VLCNs Require Connection Multiplexing," Proceedings of the IEEE ICNN, San Diego, Calif., Jul. 1983, vol. II, pp. 173-180. M. Rundnick and D. Hammerstrom, "An Interconnect Structure for Wafer Scale Neurocomputers," Abstracts of the First annual INNS Meeting, Boston, Mass., Sep. 1988, p. 405. T. Beynon and N. Dodd, "The Implementation of Multi-Layer Perceptron on Transputer Networks," Parallel Programming of Transputer Based Machines: Proceedings of the 7th OCCAM User Group Technical Meeting Grenoble, France, Sep. 1987, pp. 108-119. ART-UNIT: 238 PRIM-EXMR: Allen R. MacDonald ASST-EXMR: George Davis LEGAL-REP: Nixon & Vanderhye ABSTRACT: A neurocomputer connected to a host computer, the neurocomputer having a plurality of processor elements, each of the processor elements being placed at each of node of a lattice respectively, the neurocomputer includes a plurality of first processor elements, each of the first processor elements being placed at a node of the lattice, capable of transmitting data from and to the host computer and capable of transmitting the data to one of adjacent processor elements, a plurality of second processor elements, each of the second processor elements being placed at a node of the lattice, capable of receiving the data from one of adjacent processor elements, and capable of sending the data to another adjacent processor elements from which the data is not outputted. The neurocomputer also includes a plurality of rectangular regions, each of the rectangular regions including a plurality of the processor elements, a plurality of physical processors, each of the processors being placed in each of the rectangular regions and connected with adjacent processors each other, each of the processors being capable of inputting and outputting to and from the host computer and having all functions of the processor elements included in the rectangular region, and a device for distributing the physical processors to one or a plurality of divided sections formed in the rectangular regions in such a manner that each of the sections is substantially equally assigned to each of the physical processors by permutation. 33 Claims, 55 Drawing Figures EXMPL-CLAIM: 1 NO-PP-DRAWING: 45 ============================================================================== -- ************************************************************************** Greg Aharonian Source Translation & Optimiztion P.O. Box 404, Belmont, MA 02178