CN102023672B - Integrated optical vector-matrix multiplier based on micro-ring resonators - Google Patents
Integrated optical vector-matrix multiplier based on micro-ring resonators Download PDFInfo
- Publication number
- CN102023672B CN102023672B CN 200910093881 CN200910093881A CN102023672B CN 102023672 B CN102023672 B CN 102023672B CN 200910093881 CN200910093881 CN 200910093881 CN 200910093881 A CN200910093881 A CN 200910093881A CN 102023672 B CN102023672 B CN 102023672B
- Authority
- CN
- China
- Prior art keywords
- vector
- matrix
- matrix multiplier
- micro
- inner product
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses an integrated optical vector-matrix multiplier. The optical vector-matrix multiplier comprises extensible inner product operation units formed by a plurality of tunable micro-ring resonators which have a parallel waveguide structure and are arranged in a one-dimensional manner. The optical vector-matrix multiplier realizes a multiplication between a matrix M*N and a vector N*1 by employing the integrated optical mode, wherein elements in the matrix and the vector are all binary numbers (0 or 1) and M and N are natural numbers. The integrated optical vector-matrix multiplier based on the micro-ring resonators is high in integrated level, good in scalability and convenient in integration with electric elements, and enables to raise information transfer rates, reduce information transfer delays, avoid disadvantages of poor precision and weak programmable ability of analog optical computing systems and solve the problem of low space diversity efficiency of traditional optical vector-matrix multiplier.
Description
Technical field
The present invention relates to the optical vector-matrix multiplier technical field, relate in particular to a kind of integrated optical vector-matrix multiplier based on micro-ring resonator (Optical Vector-Matrix Multiplier, OVMM), this optical vector-matrix multiplier is arranged to realize by two dimension by the micro-ring resonator (Microring Resonator, MRR) of tunable straight parallel waveguiding structure.
Background technology
At first, the model of optical vector-matrix multiplier is that the J.W.Goodman of U.S. Stanford university proposes, " Fully parallel; high speed incoherentoptical method for performing discrete Fourier the transforms " (J.W.Goodman that delivered in 1978, OPTICS LETTERS, Vol.2, No.1,1-3) it is introduced.Goodman " Stamford Vector-Matrix Multiplier " mainly comprises three parts, i.e. array of source, matrix mask and detector array.At first utilize delegation's light emitting diode as input vector, between light emitting diode matrix and matrix mask, arrange light path to make each light emitting diode only illuminate row of matrix mask, behind matrix mask, arrange again such light path, light from the every delegation of matrix mask is focused on the different detecting elements, thereby realized 10 discrete Fourier transformation.Calculated the huge applications potentiality in field at photometry although it has been recognized that at that time OVMM, but the reason owing to technology, the precision of array of source, detector array and spatial light modulator, dimension and speed are all not ideal enough, so the work of this respect stagnates always.In addition, because light signal is in free-space propagation and diversity, capacity usage ratio is low, and system and resetting difficulty are large.These shortcomings have all limited its range of application.
The stagnation that photometry is calculated research is because the photometry calculation itself faces a lot of challenges on the one hand, on the other hand also because the development of integrated circuit.Since the seventies in last century, along with the progress of semiconductor process techniques, obtained to reach the quick and lasting development of three more than ten years based on the microprocessor of integrated circuit.We know that transistorized size is done littlely on the chip, and the frequency of CPU work just can be higher.Along this thinking, people improve the performance of the CPU prophesy according to Moore's Law again and again.But develop into today, some new problems occurred again, high primary frequency causes the power consumption of CPU sharply to rise, and electric leakage and heat dissipation problem also are difficult to solve.Transistor size integrated on the chip means that near the limit of physics the performance of integrated circuit also will reach capacity, and Moore's Law will lose efficacy soon.For this problem, people begin to consider to adopt new physical principle to realize processor and computing machine, wherein mainly comprise optical computer and quantum computer.Although based on the practical in addition segment distance of universal microprocessor distance of these concepts, the development of some essential elements or prototype machine is very meaningful.Because these new ideas computing machines adopt some special physical principles, they may be good at and process a certain generic task very much.Just has the high advantage of concurrency such as the light processing mode.This is so that perhaps they processing for some special duties provide association at present, can become microprocessor and the Computer realizing way of main flow in future.
The research of the optical vector-matrix multiplier that scientist's early starts such as the Germany Hagen M.Gruber of university are integrated, be published in " Planar-integrated optical vector-matrixmultiplier " (Matthias Gruber in 2002, APPLIED OPTICS, Vol.39, No.29,5367～5373) put down in writing them and realized the method for OVMM at semiconductor wafer.Its basic thought remains such that professor J.W.Goodman proposes, and has only adopted now the technology of semiconductor wafer.The function of lens adopts lenticule and the diffraction element (DOE, Diffraction Optical Element) of being made by lithographic technique to realize here in J.W.Goodman professor's the model.Because still adopt the mode of Free propagation (in semiconductor material) to realize the diversity of light signal, the energy efficiency of system is very low, so just very high to the requirement of optical light source and detector, this is just so that it is difficult to really practical.
The device of a kind of OVMM is disclosed in patent US 4,620,293 " Optical matrix multiplier ".Wherein, utilize light emitting diode (LED) that input vector is provided, provide the matrix of input by spatial light modulator (SLM), utilize photodetector to realize the detection of light signal, realized optical vector-matrix multiplication.As seen, it namely is the specific implementation of the model of J.W.Goodman proposition.
The device of a kind of OVMM is disclosed in patent EP Isosorbide-5-Nitrae 36,681 " Vector-Matrix Multiplication ".Wherein, utilize vertical cavity surface emitting laser (VCSEL) that input vector is provided, provide the matrix of input by Multiple Quantum Well spatial light modulator (MQW-SLM), utilize photodetector to realize the detection of light signal, realized optical vector-matrix multiplication.And algorithm, function library, storer and peripheral control unit have been designed for it.Although it has adopted more advanced device, but still it is the specific implementation of the spatial diversity model of J.W.Goodman proposition.
A kind of device of integrated optical vector-matrix multiplier is disclosed in patent US 4,125,316 " Integrated optical matrix multiplier ".It adopts photoconductive material waveguide a and electrooptical material waveguide b to make, and when light process a, causes that the electric field at b place changes, thereby the light among the b is had an effect.This is to utilize electric field to realize light for medium multiplication is finished in the modulation of light in fact.But this scheme need to be used photoconductive material and electrooptical material, and is incompatible with integrated circuit technology.
Summary of the invention
The technical matters that (one) will solve
In view of this, fundamental purpose of the present invention is to provide a kind of optical vector-matrix multiplier based on micro-ring resonator, to improve message data rate, reduce the delay that information is transmitted, avoid the shortcomings such as the low precision, programmability of simulated light computing system be weak, and solve the problem of traditional optical Vector-Matrix Multiplier space diversity inefficiency.
(2) technical scheme
For achieving the above object, the invention provides a kind of integrated optical vector-matrix multiplier, this optical vector-matrix multiplier is by the inner product operation cell formation of Parallel coupled, this inner product operation unit has the micro-ring resonator of the tunable parallel waveguide structure of a plurality of one dimensions arrangements, this optical vector-matrix multiplier adopts the mode of integrated optics to realize the multiplication of M * N matrix and N * 1 vector, wherein the element in matrix and the vector is 0 or 1 binary number, and M, N are natural number.
In the such scheme, this micro-ring resonator is the micro-ring resonator that the input and output straight wave guide is parallel to each other, rather than the micro-ring resonator of crossing waveguide structure, does not have coupling between the ring of micro-ring resonator.
In the such scheme, this micro-ring resonator comprises two straight wave guides that are parallel to each other and a disc waveguide, and this disc waveguide and two straight wave guides are tangent.
In the such scheme, the function implementation procedure of this Vector-Matrix Multiplier is:
Input end is the straight wave guide port of M inner product operation unit, and input signal is determined by initial vector, if i element of initial vector is 1, then all M port input all contains corresponding wavelength composition λ
iLaser pulse; If i element of initial vector is 0, then all the input of M port does not contain this wavelength signals;
If the capable j column element of i of matrix is 1, then the resonance wavelength of j MRR of i inner product operation unit is tuned to λ
iIf the capable j column element of i of matrix is 0, then the resonance wavelength of j MRR of i inner product operation unit is adjusted to non-λ
1, λ
2..., λ
NArbitrary wavelength place, j=1 wherein, 2 ..., N;
M output port at optical vector-matrix multiplier surveyed, and the gained light intensity is followed successively by the element of vector-matrix multiplication gained M dimension result vector.
In the such scheme, the different elements in this Vector-Matrix Multiplier input vector are the laser pulse of corresponding different wave length respectively, and when certain element in the input vector was 0, the laser pulse power of corresponding wavelength was zero; When certain element was 1, the power of the laser pulse of corresponding wavelength was a default normal value; In each computing, the input of each port of Vector-Matrix Multiplier is identical.
In the such scheme, characterize the element of input matrix with the resonant condition of MRR, the corresponding inner product operation unit of every delegation of matrix, the capable j column element of i is 1 o'clock, the resonance frequency of j little ring is λ
jWhen this element was 0, the resonance frequency of j little ring was non-λ
1, λ
2..., λ
NArbitrary wavelength, j=1 wherein, 2 ..., N.
(3) beneficial effect
Can find out that from technique scheme the present invention has following beneficial effect:
Integrated optical vector-matrix multiplier provided by the invention utilizes ready-made technology, and is low in energy consumption so that device volume is little, and favorable expandability is convenient to electricity component integrated; Utilize the laser pulse transmission of information, speed is high, postpones little; Carry out signal with digital form and process, avoided low precision, the weak shortcoming of programmability of simulated light computing system; With high index-contrast waveguide light conducting, solved the problem of traditional optical Vector-Matrix Multiplier space diversity inefficiency.
Description of drawings
Fig. 1 is the micro-ring resonator of parallel waveguide structure;
Fig. 2 is the inner product operation unit;
Fig. 3 is the structural representation based on the integrated optical vector-matrix multiplier of micro-ring resonator.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, below in conjunction with specific embodiment, and with reference to accompanying drawing, the present invention is described in more detail.
Fig. 1 is the elementary cell that consists of optical vector-matrix multiplier, the i.e. tunable micro-ring resonator of straight parallel waveguiding structure.Port one is input end, and port 2 is straight-through end, and port 3 is output terminal, and supposes that the resonance frequency of little ring is λ
1Utilize single tunable micro-ring resonator can realize the multiplication of two bits, the input/output end port that it is corresponding and the resonant condition of little ring are as shown in table 1, and table 1 is to utilize single micro-ring resonator to realize the multiplication of two binary numbers.Wherein, the input light of port one and the resonance wavelength of little ring represent respectively two multipliers, and namely port one has λ
1Light when input, first multiplier is 1, otherwise is 0, the micro-ring resonant wavelength is λ
1The time, second multiplier is 1, otherwise is 0; Result of calculation is represented have the light time to represent that the result is 1, otherwise be 0 by the output of port 3.
Table 1
Fig. 2 is the inner product operation unit, can realize the inner product operation of two binary vectors.What realize such as Fig. 2 is inner product operation between the vector [1 01 1] and [0 11 1].Detailed process is: be respectively λ from the input port input wavelength
1, λ
3, λ
4Light (expression vector [1 01 1]); Modulate simultaneously the 1st little ring, making its resonance wavelength is non-λ
1, λ
2, λ
3Or λ
4Any wavelength, and the resonance wavelength of other little rings is constant, is followed successively by λ
2, λ
3, λ
4(representing another vector [0 11 1]); Then will to comprise wavelength be λ to output terminal
3And λ
4Light, operation result (luminous power of output terminal) records by the photodetector of output terminal.
Fig. 3 is the structural representation based on the integrated optical vector-matrix multiplier of micro-ring resonator.On the basis of the principle of work of having understood inner product operation unit shown in Figure 2, be not difficult to find out that the principle of this scheme witness vector and matrix multiplication operation is very directly perceived.So-called matrix-vector multiplier only need to repeat to make a plurality of inner product operations unit and get final product, and when OVMM worked: synchronization was inputted the vector of all inner product operation devices, and namely laser pulse combination all is identical; The resonance wavelength of different is all little rings all is independently to control according to matrix to be calculated at that time.
Following formula is the general expression of this optical vector-matrix multiplier M that can finish * N matrix and N * 1 vector multiplication computing.Formula above below contrasting, Fig. 3 is done the explanation of some necessity:
1) input of four input ports of synchronization all is vectorial X=[x
1x
2x
3x
4], vector element x
iBe 1/0, λ in the corresponding input light wave
iComposition with/without;
2) row of the row of matrix/corresponding little ring array of row difference/row.As arbitrary matrix element w
Ij=1 o'clock, the resonance wavelength of the micro-ring resonator of the capable j row of corresponding i was λ
jWork as w
Ij=0 o'clock, the resonance wavelength of the micro-ring resonator of the capable j row of corresponding i was non-λ
1, λ
2..., λ
NArbitrary wavelength;
3) each inner product operation unit, the delegation of realization matrix and the inner product operation of input vector obtain an element in the input vector.The multiplying of M * N matrix and N * 1 vector just can be realized in M such unit side by side.
The above specifies with regard to the principle of work of optical vector-matrix multiplier, and needing more ben is the setting of resonance wavelength.The λ that we mention in front
1, λ
2..., λ
NAnd non-λ
1, λ
2..., λ
NArbitrary wavelength all be to be limited in the same free spectral limit (Free Spectral Range, FSR).This is to be determined by the filtering characteristic of micro-ring resonant wave filter, supposes that N little ring in the inner product operation device has identical FSR and three dB bandwidth, then λ
1, λ
2..., λ
NCan set according to two three dB bandwidths of being separated by successively, so just require FSR to be at least 2N three dB bandwidth doubly, when needing to change 0,1 state of little ring, only the resonance wavelength of little ring need to be regulated a three dB bandwidth and get final product.This structure have one outstanding be a little to have good extendability: on the one hand, because the loss of each inner product operation unit only comes the loss in the self-waveguide, so little ring that can a lot of different resonance wavelengths of serial arrangement, namely the dimension N of vector can get very large; On the other hand, and the line number that increases matrix only means and increase identical inner product operation unit that the value of M is restricted hardly.
Optical vector matrix multiplier based on the parallel ring resonator provided by the invention can be made of different materials, and such as SOI, SiN and III-V family material etc., tuning mode also has multiple choices, and principle is directly perceived, realizes that simply extensibility is strong.All little rings all adopt identical design and processes flow process, just the MRR of different lines has intrinsic separately resonance wavelength in the matrix, and this can be by realizing the different physical dimension (comprising the spacing of radius, straight wave guide and the ring of little ring, the sectional dimension of disc waveguide etc.) of each MRR design.
Above-described specific embodiment; purpose of the present invention, technical scheme and beneficial effect are further described; institute is understood that; the above only is specific embodiments of the invention; be not limited to the present invention; within the spirit and principles in the present invention all, any modification of making, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (4)
1. integrated optical vector-matrix multiplier, it is characterized in that, this optical vector-matrix multiplier is by a plurality of extendible inner product operation cell formation of Parallel coupled, each inner product operation unit is made of the micro-ring resonator of a plurality of tunable parallel waveguide structure that one dimension is arranged, and this optical vector-matrix multiplier is arranged to realize by two dimension by the micro-ring resonator of tunable straight parallel waveguiding structure; This optical vector-matrix multiplier adopts the mode of integrated optics to realize the multiplication of M * N matrix and N * 1 vector, and wherein the element in matrix and the vector is 0 or 1 binary number, and M, N are natural number;
The micro-ring resonator of described parallel waveguide structure is the micro-ring resonator that the input and output straight wave guide is parallel to each other, rather than the micro-ring resonator of crossing waveguide structure, there is not coupling between the ring of micro-ring resonator, comprise two straight wave guides that are parallel to each other and a disc waveguide, this disc waveguide and two straight wave guides are tangent, and two straight wave guides are respectively input straight wave guide, straight wave guide output;
The specific works process of this optical vector-matrix multiplier is:
The light signal that at first, will represent N * 1 vector is input in M the inner product operation unit simultaneously;
In the light signal of input expression N * 1 vector, the M in M * N matrix 1 * N vector is loaded into respectively M inner product operation unit, wherein a micro-ring resonator in each the control of element inner product operation unit in 1 * N vector;
When the light signal of expression N * 1 vector passes through this optical vector-matrix multiplier, the capable multiplying with N * 1 vector of i in M * N matrix will be finished in i inner product operation unit, and at its output terminal with i element in the formal output M of light intensity * 1 result vector, i=1 wherein, 2, ..., M;
This optical vector-matrix multiplier comprises M inner product operation unit altogether, thereby can finish the multiplying of M * N matrix and N * 1 vector.
2. integrated optical vector-matrix multiplier according to claim 1 is characterized in that, i element corresponding wavelength of this N * 1 vector is λ
iLaser pulse, i=1 wherein, 2 ..., N; When i element of input vector was 0, wavelength was λ
iThe power of laser pulse be zero; When i element was 1, wavelength was λ
iThe power of laser pulse be a default normal value.
3. integrated optical vector-matrix multiplier according to claim 1, it is characterized in that, different elements in this Vector-Matrix Multiplier input vector are the laser pulse of corresponding different wave length respectively, and when certain element in the input vector was 0, the laser pulse power of corresponding wavelength was zero; When certain element was 1, the power of the laser pulse of corresponding wavelength was a default normal value; In each computing, the input of each port of Vector-Matrix Multiplier is identical.
4. integrated optical vector-matrix multiplier according to claim 1, it is characterized in that, this optical vector-matrix multiplier characterizes the element of input matrix with the resonant condition of micro-ring resonator in the inner product operation unit, M * N matrix comprises the row vector of M 1 * N, each corresponding inner product operation unit of row vector, when the capable j column element of i was 1 in M * N matrix, the resonance frequency of j little ring was λ in i inner product operation unit
jWhen this element was 0, the resonance frequency of j micro-ring resonator was non-λ in i inner product operation unit
1, λ
2..., λ
NArbitrary wavelength, i=1 wherein, 2 ..., M, j=1,2 ..., N.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910093881 CN102023672B (en) | 2009-09-23 | 2009-09-23 | Integrated optical vector-matrix multiplier based on micro-ring resonators |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910093881 CN102023672B (en) | 2009-09-23 | 2009-09-23 | Integrated optical vector-matrix multiplier based on micro-ring resonators |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102023672A CN102023672A (en) | 2011-04-20 |
| CN102023672B true CN102023672B (en) | 2013-03-20 |
Family
ID=43865058
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910093881 Active CN102023672B (en) | 2009-09-23 | 2009-09-23 | Integrated optical vector-matrix multiplier based on micro-ring resonators |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102023672B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11853871B2 (en) | 2018-06-05 | 2023-12-26 | Lightelligence PTE. Ltd. | Optoelectronic computing systems |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103322904B (en) * | 2013-06-09 | 2015-04-22 | 中国科学院长春光学精密机械与物理研究所 | Optical heterodyne interference method for eliminating non-linear errors based on resonant filtering method and inner product method |
| CN103473213A (en) * | 2013-09-12 | 2013-12-25 | 中国科学院半导体研究所 | System for loading and extracting parallel information of optical vector-matrix multiplier |
| CN103678258B (en) * | 2013-12-25 | 2017-01-25 | 中国科学院半导体研究所 | Method for improving data resolution ratio of silica-based optical matrix processor |
| CN104764531B (en) * | 2015-03-13 | 2017-10-13 | 深圳先进技术研究院 | Integrated infrared heat sensor and its manufacture method and imaging system and imaging method |
| CN107832640B (en) * | 2017-10-25 | 2021-07-06 | 上海交通大学 | Normalized non-negative real matrix convolver and method of use |
| CN107977231B (en) * | 2017-12-15 | 2020-10-27 | 安徽寒武纪信息科技有限公司 | Calculation method and related product |
| TWI735886B (en) | 2018-06-05 | 2021-08-11 | 美商光子智能股份有限公司 | Computing system |
| CN109188825A (en) * | 2018-10-09 | 2019-01-11 | 宁波大学 | Optics half adder based on graphene surface plasmon |
| US11734556B2 (en) | 2019-01-14 | 2023-08-22 | Lightelligence PTE. Ltd. | Optoelectronic computing systems |
| CN110703851B (en) * | 2019-11-04 | 2021-06-22 | 兰州大学 | Optical matrix vector multiplier based on mode multiplexing |
| US11719963B2 (en) | 2020-04-29 | 2023-08-08 | Lightelligence, Inc. | Optical modulation for optoelectronic processing |
| CN111561953B (en) * | 2020-05-14 | 2021-11-05 | 上海科技大学 | On-chip integrated optical matrix vector multiplier based on wavelength division multiplexing and balanced detection |
| CN112799464B (en) * | 2021-04-01 | 2021-07-27 | 中国科学院西安光学精密机械研究所 | Matrix vector multiplier based on multi-wavelength light source and operation method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4620293A (en) * | 1983-12-23 | 1986-10-28 | General Dynamics, Pomona Division | Optical matrix multiplier |
| US5432722A (en) * | 1987-03-27 | 1995-07-11 | Opticomp Corporation | Global interconnect architecture for electronic computing modules |
| CN2323388Y (en) * | 1997-12-27 | 1999-06-09 | 中国科学院长春物理研究所 | Integrated optical acousto-optic array multiplier assembly |
| CN102147634A (en) * | 2010-11-24 | 2011-08-10 | 中国科学院半导体研究所 | Optical vector-matrix multiplier based on single-waveguide coupling micro-ring resonant cavity |
-
2009
- 2009-09-23 CN CN 200910093881 patent/CN102023672B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4620293A (en) * | 1983-12-23 | 1986-10-28 | General Dynamics, Pomona Division | Optical matrix multiplier |
| US5432722A (en) * | 1987-03-27 | 1995-07-11 | Opticomp Corporation | Global interconnect architecture for electronic computing modules |
| CN2323388Y (en) * | 1997-12-27 | 1999-06-09 | 中国科学院长春物理研究所 | Integrated optical acousto-optic array multiplier assembly |
| CN102147634A (en) * | 2010-11-24 | 2011-08-10 | 中国科学院半导体研究所 | Optical vector-matrix multiplier based on single-waveguide coupling micro-ring resonant cavity |
Non-Patent Citations (5)
| Title |
|---|
| 冷月华.硅基集成化电光调制器的研究.《中国优秀硕士学位论文全文数据库信息科技辑》.2011,(第4期),26-40,64. * |
| 吴晖等.一种新颖的可重组乘法器设计.《微电子学与计算机》.2004,第21卷(第9期),161-163. * |
| 张锐等.一套二进制向量_矩阵乘法运算的光计算系统.《计算机技术与发展》.2007,第17 卷(第10 期),141-143. * |
| 张锐等.光学向量矩阵乘法器的进展.《仪器仪表学报》.2006,第27卷(第6期),960-962. * |
| 李修建等.光学向量_矩阵乘法器教学演示实验.《大学物理》.2008,第27卷(第8期),38-41. * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11853871B2 (en) | 2018-06-05 | 2023-12-26 | Lightelligence PTE. Ltd. | Optoelectronic computing systems |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102023672A (en) | 2011-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102023672B (en) | Integrated optical vector-matrix multiplier based on micro-ring resonators | |
| Liu et al. | Holylight: A nanophotonic accelerator for deep learning in data centers | |
| Ahn et al. | Devices and architectures for photonic chip-scale integration | |
| CN102866876B (en) | Single chip integrated optical matrix-vector multiplier | |
| CN114358271B (en) | Time-wavelength interweaving photon neural network convolution acceleration chip | |
| CN101872039A (en) | 4*4 non-blocking optical router based on active micro-ring resonator | |
| CN113890620B (en) | Silicon substrate photonic neural network based on tunable filter and modulation method thereof | |
| CN113961035A (en) | Deep learning-oriented high-efficiency reconfigurable all-optical neural network computing chip architecture | |
| Peng et al. | Dnnara: A deep neural network accelerator using residue arithmetic and integrated photonics | |
| CN108182049B (en) | Binary optical full adder based on micro-ring resonator | |
| Kononov | Modeling photonic links in Verilog-A | |
| Yao et al. | Thermal-sensitive design and power optimization for a 3D torus-based optical NoC | |
| Hossain et al. | Speed enhancement of all-optical pseudo random binary sequence (PRBS) generator using microring resonator | |
| CN102147634A (en) | Optical vector-matrix multiplier based on single-waveguide coupling micro-ring resonant cavity | |
| Li et al. | An energy-efficient silicon photonic-assisted deep learning accelerator for big data | |
| Yang et al. | Inter/intra-chip optical interconnection network: Opportunities, challenges, and implementations | |
| CN113935493A (en) | Programmable high-dimensional quantum computing chip structure based on integrated optics | |
| Sunny et al. | Machine learning accelerators in 2.5 d chiplet platforms with silicon photonics | |
| Zhu et al. | Dota: A dynamically-operated photonic tensor core for energy-efficient transformer accelerator | |
| CN111563583B (en) | Optical computing chip design method based on neural ordinary differential equation and optical computing chip | |
| Wu et al. | Pairing of microring-based silicon photonic transceivers for tuning power optimization | |
| Shafiee et al. | Inexact silicon photonics: From devices to applications | |
| Peserico et al. | FFT-based Convolution Neural Network on Silicon Photonics Platform | |
| CN103595419A (en) | N-bit binary-system electro-optic odd-even checker | |
| Xu et al. | High-energy-efficiency integrated photonic convolutional neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| ASS | Succession or assignment of patent right |
Owner name: HUAWEI TECHNOLOGY CO., LTD. Free format text: FORMER OWNER: SEMICONDUCTOR INST., CHINESE ACADEMY OF SCIENCES Effective date: 20131209 |
|
| C41 | Transfer of patent application or patent right or utility model | ||
| COR | Change of bibliographic data |
Free format text: CORRECT: ADDRESS; FROM: 100083 HAIDIAN, BEIJING TO: 518129 SHENZHEN, GUANGDONG PROVINCE |
|
| TR01 | Transfer of patent right |
Effective date of registration: 20131209 Address after: 518129 Bantian HUAWEI headquarters office building, Longgang District, Guangdong, Shenzhen Patentee after: Huawei Technologies Co., Ltd. Address before: 100083 Beijing Qinghua East Road, Haidian District, No. 35 Patentee before: Semiconductor Inst., Chinese Academy of Sciences |