Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.tud1
A. Krishnamoorthy, J. Ford, G. Marsden, G. Yayla, S. Esener
A variety of applications in artificial neural networks, interconnection networks, artificial intelligence, relational databases, and numerical processing require parallel, large scale implementations of matrix-algebraic architectures. Existing VLSI implementations of these architectures are restricted in terms of their parallelism and bandwidth due to their inherent connectivity, pin-out, power dissipation, and crosstalk limitations.[1,2] On the other hand, existing optical matrix-vector architectures suffer from limited SLM throughput and accuracy as well as limited functional flexibility. In the following sections we describe and analyze the Dual-Scale Topology OptoElectronic Processor (D-STOP)[3] which alleviates these limitations, and discuss its feasibility for a near-term implementation.
{"title":"Dual Scale Topology Opto-Electronic Processor (D-STOP): Comparative Analysis and Technological Feasibility","authors":"A. Krishnamoorthy, J. Ford, G. Marsden, G. Yayla, S. Esener","doi":"10.1364/optcomp.1991.tud1","DOIUrl":"https://doi.org/10.1364/optcomp.1991.tud1","url":null,"abstract":"A variety of applications in artificial neural networks, interconnection networks, artificial intelligence, relational databases, and numerical processing require parallel, large scale implementations of matrix-algebraic architectures. Existing VLSI implementations of these architectures are restricted in terms of their parallelism and bandwidth due to their inherent connectivity, pin-out, power dissipation, and crosstalk limitations.[1,2] On the other hand, existing optical matrix-vector architectures suffer from limited SLM throughput and accuracy as well as limited functional flexibility. In the following sections we describe and analyze the Dual-Scale Topology OptoElectronic Processor (D-STOP)[3] which alleviates these limitations, and discuss its feasibility for a near-term implementation.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114443854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.me2
Jian Ma, Y. Taketomi, Y. Fainman, J. Ford, Sing H. Lee
Photorefractive (PR) devices have found applications in optical computing, image processing and pattern recognition[1–3], because PR materials provide unique features such as real time operation, optical gain, storage, nonlinear operations, phase conjugation and correlation. New PR materials are being investigated in order to meet the device and system requirements of sensitivity, speed, and operation wavelength (e.g., response to the near infrared spectral range for systems operated with semiconductor lasers). Compound semiconductors may satisfy these requirements. For example, optical signal amplification by two-beam coupling and amplified phase-conjugate beam reflection by four-wave mixing have been reported in GaAs[4] and InP[5] at the wavelength of 1.06 μm. Recently, GaP[6–7] was shown to possess a relatively weak PR effect in the spectral range of 0.6 to 0.9 μm. In this manuscript we report enhancement of the PR effect in GaP using an externally applied electric field and moving grating. In particular, two- and four-wave mixing experiments were used to demonstrate a gain coefficient of Γ = 1.9 cm–1 and a phase conjugate reflectivity, R= 4.5%. In addition, several figures of merit of GaP, i.e., steady-state index change, absorption coefficient, response time and PR sensitivity were characterized.
{"title":"Enhanced photorefractive effects with a de field and moving grating in GaP at 633 nm","authors":"Jian Ma, Y. Taketomi, Y. Fainman, J. Ford, Sing H. Lee","doi":"10.1364/optcomp.1991.me2","DOIUrl":"https://doi.org/10.1364/optcomp.1991.me2","url":null,"abstract":"Photorefractive (PR) devices have found applications in optical computing, image processing and pattern recognition[1–3], because PR materials provide unique features such as real time operation, optical gain, storage, nonlinear operations, phase conjugation and correlation. New PR materials are being investigated in order to meet the device and system requirements of sensitivity, speed, and operation wavelength (e.g., response to the near infrared spectral range for systems operated with semiconductor lasers). Compound semiconductors may satisfy these requirements. For example, optical signal amplification by two-beam coupling and amplified phase-conjugate beam reflection by four-wave mixing have been reported in GaAs[4] and InP[5] at the wavelength of 1.06 μm. Recently, GaP[6–7] was shown to possess a relatively weak PR effect in the spectral range of 0.6 to 0.9 μm. In this manuscript we report enhancement of the PR effect in GaP using an externally applied electric field and moving grating. In particular, two- and four-wave mixing experiments were used to demonstrate a gain coefficient of Γ = 1.9 cm–1 and a phase conjugate reflectivity, R= 4.5%. In addition, several figures of merit of GaP, i.e., steady-state index change, absorption coefficient, response time and PR sensitivity were characterized.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"110 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117216060","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.ma2
A. Lentine, L. Chirovsky, M. Focht, J. Freund, G. Guth
Arrays of symmetric self electro-optic effect devices (S-SEEDs) have been made with low operating energies and fast switching speeds [1,2]. The device has the characteristics of a set-reset latch, although it can be made to do logic functions such as a NOR gate by presetting the state of the device before the application of the data inputs [3]. Logic gates that can perform more complex functions without preset beams may be realized by using electrically connected detectors configured like transistors in NMOS or CMOS circuits together with an output S-SEED to provide the output beams [4]. In this paper, we describe the first integrated arrays of these logic gates, each of which can perform the four basic logic functions without the use of preset beams. Each logic gate in the array consists of six quantum well p-i-n diodes, four input diodes configured similar to transistors in a CMOS NOR gate, and two output diodes (i. e. a S-SEED) that provide a set of complementary output beams. Like the S-SEEDs, this device has time sequential gain, in which the low power input beams set the state of the device and a set of equal higher power clock beams subsequently read the state. This device retains many desirable qualities of the S-SEED such as signal regeneration and retiming, wavefront restoration, and operation over several decades in power levels due to its differential nature. Because the logic gate contains only quantum well diodes, the same batch fabrication procedures [1] used for S-SEED arrays were used to make the arrays of these devices.
{"title":"Integrated array of self electro-optic effect device logic gates","authors":"A. Lentine, L. Chirovsky, M. Focht, J. Freund, G. Guth","doi":"10.1364/optcomp.1991.ma2","DOIUrl":"https://doi.org/10.1364/optcomp.1991.ma2","url":null,"abstract":"Arrays of symmetric self electro-optic effect devices (S-SEEDs) have been made with low operating energies and fast switching speeds [1,2]. The device has the characteristics of a set-reset latch, although it can be made to do logic functions such as a NOR gate by presetting the state of the device before the application of the data inputs [3]. Logic gates that can perform more complex functions without preset beams may be realized by using electrically connected detectors configured like transistors in NMOS or CMOS circuits together with an output S-SEED to provide the output beams [4]. In this paper, we describe the first integrated arrays of these logic gates, each of which can perform the four basic logic functions without the use of preset beams. Each logic gate in the array consists of six quantum well p-i-n diodes, four input diodes configured similar to transistors in a CMOS NOR gate, and two output diodes (i. e. a S-SEED) that provide a set of complementary output beams. Like the S-SEEDs, this device has time sequential gain, in which the low power input beams set the state of the device and a set of equal higher power clock beams subsequently read the state. This device retains many desirable qualities of the S-SEED such as signal regeneration and retiming, wavefront restoration, and operation over several decades in power levels due to its differential nature. Because the logic gate contains only quantum well diodes, the same batch fabrication procedures [1] used for S-SEED arrays were used to make the arrays of these devices.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128978219","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.mc1
A. Lohmann, S. Sinzinger
An ideal array illuminator would provide equal amounts of light power to all elements of an array of gates or smart pixels. Existing array illuminators (abbreviated: AIL) achieve a homogeneity of 5 to 10%. That may seem to be good enough, if the signals are binary. However, it is desirable to achieve the best possible homogeneity, since there might be other causes for inhomogeneous behavior of the array system. Having a good AIL relieves the burden of tolerances for the other components of the overall system. Furthermore, when the signals are analog, as in some neural systems, the homogeneity of the power supply becomes even more important.
{"title":"Spatial Noise Reduction in Array Illuminators","authors":"A. Lohmann, S. Sinzinger","doi":"10.1364/optcomp.1991.mc1","DOIUrl":"https://doi.org/10.1364/optcomp.1991.mc1","url":null,"abstract":"An ideal array illuminator would provide equal amounts of light power to all elements of an array of gates or smart pixels. Existing array illuminators (abbreviated: AIL) achieve a homogeneity of 5 to 10%. That may seem to be good enough, if the signals are binary. However, it is desirable to achieve the best possible homogeneity, since there might be other causes for inhomogeneous behavior of the array system. Having a good AIL relieves the burden of tolerances for the other components of the overall system. Furthermore, when the signals are analog, as in some neural systems, the homogeneity of the power supply becomes even more important.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126420138","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.me12
D. Lu, Ting-Ting Y. Lin, F. Kiamilev, S. Esener, Sing H. Lee
Wafer scale integration (WSI) promises to realize a complete multiprocessing system on the same wafer and eliminates the expensive steps required to dice and bond. The fundamental belief is that the internal connection between chips on the same wafer are more reliable and have a smaller propagation delay than external connections1. However, achieving a high yield has proven to be a major challenge. Rather than aiming for 100% yield, the realistic solution is to determine the defective components on the wafer and replace them with spares. Which means, the design should be tolerant to faults developed during the manufacturing process. Moreover, faults occur during system operation, be it component failure, improper operation, or environmental factors. Therefore, a mean to detect these unexpected faults and recover from them is necessary to minimize down time and unavailability. Long and periodic system downs are a luxury that cannot be afforded for computers used in critical applications. In this paper, we show that the introduction of optical interconnection techniques into a multiprocessor environment (e.g. the Programmable Optoelectronic Multiprocessor, POEM) enables efficient implementation of fault-tolerant techniques.
{"title":"Fault-Tolerant Computing on POEM","authors":"D. Lu, Ting-Ting Y. Lin, F. Kiamilev, S. Esener, Sing H. Lee","doi":"10.1364/optcomp.1991.me12","DOIUrl":"https://doi.org/10.1364/optcomp.1991.me12","url":null,"abstract":"Wafer scale integration (WSI) promises to realize a complete multiprocessing system on the same wafer and eliminates the expensive steps required to dice and bond. The fundamental belief is that the internal connection between chips on the same wafer are more reliable and have a smaller propagation delay than external connections1. However, achieving a high yield has proven to be a major challenge. Rather than aiming for 100% yield, the realistic solution is to determine the defective components on the wafer and replace them with spares. Which means, the design should be tolerant to faults developed during the manufacturing process. Moreover, faults occur during system operation, be it component failure, improper operation, or environmental factors. Therefore, a mean to detect these unexpected faults and recover from them is necessary to minimize down time and unavailability. Long and periodic system downs are a luxury that cannot be afforded for computers used in critical applications. In this paper, we show that the introduction of optical interconnection techniques into a multiprocessor environment (e.g. the Programmable Optoelectronic Multiprocessor, POEM) enables efficient implementation of fault-tolerant techniques.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127001320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-22DOI: 10.1364/optcomp.1991.mc3
H. Ozaktas, J. Goodman
A major advantage of optical and superconducting interconnections is their ability to transfer large amounts of information per unit cross section over long distances. Let the maximum information flux a given communication medium can support be denoted by I and be measured in bits/m 2 sec. For the length scales involved in a computing system (< 10 m), it is possible to reduce the effects of dispersion and attenuation to the extent that I may be assumed to be independent of length for optical and superconducting interconnections. On the other hand, 1 is a decreasing function of communication length for resistive interconnections, making them disadvantageous over longer distances. However, for distances less than about the order of a centimeter, they can provide greater information flux than optical or superconducting interconnections.
{"title":"Multiplexed Hybrid Interconnection Architectures","authors":"H. Ozaktas, J. Goodman","doi":"10.1364/optcomp.1991.mc3","DOIUrl":"https://doi.org/10.1364/optcomp.1991.mc3","url":null,"abstract":"A major advantage of optical and superconducting interconnections is their ability to transfer large amounts of information per unit cross section over long distances. Let the maximum information flux a given communication medium can support be denoted by I and be measured in bits/m\u0000 2\u0000 sec. For the length scales involved in a computing system (< 10 m), it is possible to reduce the effects of dispersion and attenuation to the extent that I may be assumed to be independent of length for optical and superconducting interconnections. On the other hand, 1 is a decreasing function of communication length for resistive interconnections, making them disadvantageous over longer distances. However, for distances less than about the order of a centimeter, they can provide greater information flux than optical or superconducting interconnections.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128600044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-05-12DOI: 10.1364/optcomp.1991.wa4
M. Neifeld, D. Psaltis
The optical disk is a simple computer addressable binary storage medium with very high capacity. [1] More than 1010 bits of information can be recorded on a 12cm diameter optical disk. The natural two dimensional format of the data recorded on optical disk makes this media particularly attractive for the storage of images and holograms, while parallel access provides a convenient mechanism through which such data may be retrieved. Parallel access to data stored on optical disk has been shown to provide interesting solutions to problems in neural networks, database retrieval and pattern recognition[2] In this paper we will discuss a closed loop associative optical memory based on the optical disk. When presented with a partial or noisy version of one of the images stored on the optical disk, the optical system evolves to a stable state in which those stored images which best match the input are temporally locked in the loop.
{"title":"Closed Loop Optical Disk Based Associative Memory","authors":"M. Neifeld, D. Psaltis","doi":"10.1364/optcomp.1991.wa4","DOIUrl":"https://doi.org/10.1364/optcomp.1991.wa4","url":null,"abstract":"The optical disk is a simple computer addressable binary storage medium with very high capacity.\u0000 [1]\u0000 More than 1010 bits of information can be recorded on a 12cm diameter optical disk. The natural two dimensional format of the data recorded on optical disk makes this media particularly attractive for the storage of images and holograms, while parallel access provides a convenient mechanism through which such data may be retrieved. Parallel access to data stored on optical disk has been shown to provide interesting solutions to problems in neural networks, database retrieval and pattern recognition[2] In this paper we will discuss a closed loop associative optical memory based on the optical disk. When presented with a partial or noisy version of one of the images stored on the optical disk, the optical system evolves to a stable state in which those stored images which best match the input are temporally locked in the loop.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125937156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1992-01-03DOI: 10.1364/optcomp.1995.otub1
B. K. Jenkins, A. Tanguay
Several broad classes of neural networks comprise distributed, nonlinear, dynamical systems in which large numbers of relatively simple processing elements (neuron units) are densely interconnected. The interconnections are often configured such that the interconnection weights are adaptive and contain the learned memories and behaviors of the system. Advanced optical interconnection techniques are being developed that can potentially be used in conjunction with optoelectronic neuron units to implement photonic neural-like computational modules (e.g., Fig. 1) with relatively large array sizes (105 to 106 neuron units) and a high degree of connectivity (fan-outs and fan-ins of 104 to 106, with 109 to 1012 total interconnections). A key open question is whether the high bandwidths (potentially 100 MHz or more) available from hybrid optoelectronic spatial light modulators (SLMs) can be effectively combined with such high density volume holographic optical interconnections (dynamically recorded in photorefractive materials) to provide enhanced computational throughput capacity as well as complex neural network simulation capability. A second key open question is whether advanced electronic/photonic packaging technologies can provide capability for system-level integration of highly compact multichip modules that exhibit both local (multi-plane) and global interconnections (Fig. 2).
{"title":"Photonic Implementations of Neural Networks","authors":"B. K. Jenkins, A. Tanguay","doi":"10.1364/optcomp.1995.otub1","DOIUrl":"https://doi.org/10.1364/optcomp.1995.otub1","url":null,"abstract":"Several broad classes of neural networks comprise distributed, nonlinear, dynamical systems in which large numbers of relatively simple processing elements (neuron units) are densely interconnected. The interconnections are often configured such that the interconnection weights are adaptive and contain the learned memories and behaviors of the system. Advanced optical interconnection techniques are being developed that can potentially be used in conjunction with optoelectronic neuron units to implement photonic neural-like computational modules (e.g., Fig. 1) with relatively large array sizes (105 to 106 neuron units) and a high degree of connectivity (fan-outs and fan-ins of 104 to 106, with 109 to 1012 total interconnections). A key open question is whether the high bandwidths (potentially 100 MHz or more) available from hybrid optoelectronic spatial light modulators (SLMs) can be effectively combined with such high density volume holographic optical interconnections (dynamically recorded in photorefractive materials) to provide enhanced computational throughput capacity as well as complex neural network simulation capability. A second key open question is whether advanced electronic/photonic packaging technologies can provide capability for system-level integration of highly compact multichip modules that exhibit both local (multi-plane) and global interconnections (Fig. 2).","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1992-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130606855","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-11-01DOI: 10.1364/optcomp.1991.me10
Soo-Young Lee, H. J. Lee, Sang-Yung Shin
After the first demonstration of optically-implemented Hopfield model [1] many neural network models have been investigated for large-scale optical implementation [2-8]. The 1-dimensional Hopfield model had been extended for 2-dimensional patterns [2], and optical implementation of bidirectional associative memory (BAM) [3-5] and quadratic associative memory [6,7] had been investigated. Adaptive neural network models such as multi-layer perceptron [8] had also been demonstrated. However performance of the simple Hopfield model and BAM is very limited, and many adaptive learning algorithms are too complicated to be implemented efficiently by optics. Also, when a new pattern need be added to the existing system, the correlation matrix learning rule of both the Hopfield model and BAM requires simple addition to existing interconnection weights, while error back-propagation learning rule for multi-layer perceptron requires to bring over all the previously stored patterns. Recently we had extended the BAM into multi-layer architecture, of which performance is quite comparable to that of multi-layer perceptron [9]. This multi-layer BAM (MBAM) still utilizes correlation matrices for easy optical implementation with outer-product matrix formation or inner-product recall. In this paper optical system architectures for the MBAM are presented for 2-dimensional patterns, and several implementation issues are discussed.
{"title":"Optical Modular Architectures for Multi-Layer Bam with 2-Dimensional Patterns","authors":"Soo-Young Lee, H. J. Lee, Sang-Yung Shin","doi":"10.1364/optcomp.1991.me10","DOIUrl":"https://doi.org/10.1364/optcomp.1991.me10","url":null,"abstract":"After the first demonstration of optically-implemented Hopfield model [1] many neural network models have been investigated for large-scale optical implementation [2-8]. The 1-dimensional Hopfield model had been extended for 2-dimensional patterns [2], and optical implementation of bidirectional associative memory (BAM) [3-5] and quadratic associative memory [6,7] had been investigated. Adaptive neural network models such as multi-layer perceptron [8] had also been demonstrated. However performance of the simple Hopfield model and BAM is very limited, and many adaptive learning algorithms are too complicated to be implemented efficiently by optics. Also, when a new pattern need be added to the existing system, the correlation matrix learning rule of both the Hopfield model and BAM requires simple addition to existing interconnection weights, while error back-propagation learning rule for multi-layer perceptron requires to bring over all the previously stored patterns. Recently we had extended the BAM into multi-layer architecture, of which performance is quite comparable to that of multi-layer perceptron [9]. This multi-layer BAM (MBAM) still utilizes correlation matrices for easy optical implementation with outer-product matrix formation or inner-product recall. In this paper optical system architectures for the MBAM are presented for 2-dimensional patterns, and several implementation issues are discussed.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"131 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1991-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127288208","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1991-03-05DOI: 10.1364/optcomp.1991.wb2
K. Kyuma, Y. Nitta, J. Ohta, S. Tai, Masanobu Takahashi
Recently, there has been a strong interest in artificial neural networks for real time applications. Among several approaches, opto-electronic neural networks1) are quite attractive because of a dense-interconnection, a parallelprocessing, and a large-scale integration capabilities using the advanced GaAs semiconductor technologies.
{"title":"The First Demonstration of an Optical Learning Chip","authors":"K. Kyuma, Y. Nitta, J. Ohta, S. Tai, Masanobu Takahashi","doi":"10.1364/optcomp.1991.wb2","DOIUrl":"https://doi.org/10.1364/optcomp.1991.wb2","url":null,"abstract":"Recently, there has been a strong interest in artificial neural networks for real time applications. Among several approaches, opto-electronic neural networks1) are quite attractive because of a dense-interconnection, a parallelprocessing, and a large-scale integration capabilities using the advanced GaAs semiconductor technologies.","PeriodicalId":302010,"journal":{"name":"Optical Computing","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1991-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115334996","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：