CN118246503A - Large-scale reconfigurable three-dimensional integrated optical neural network - Google Patents
Large-scale reconfigurable three-dimensional integrated optical neural network Download PDFInfo
- Publication number
- CN118246503A CN118246503A CN202410284681.2A CN202410284681A CN118246503A CN 118246503 A CN118246503 A CN 118246503A CN 202410284681 A CN202410284681 A CN 202410284681A CN 118246503 A CN118246503 A CN 118246503A
- Authority
- CN
- China
- Prior art keywords
- optical
- reconfigurable
- neural network
- multiplied
- layer
- 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.)
- Pending
Links
- 230000003287 optical effect Effects 0.000 title claims abstract description 136
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 54
- 238000010168 coupling process Methods 0.000 claims abstract description 30
- 238000005859 coupling reaction Methods 0.000 claims abstract description 30
- 230000008878 coupling Effects 0.000 claims abstract description 29
- 230000005540 biological transmission Effects 0.000 claims abstract description 18
- 238000001514 detection method Methods 0.000 claims abstract description 14
- 239000011159 matrix material Substances 0.000 claims description 17
- 239000012782 phase change material Substances 0.000 claims description 9
- 238000003491 array Methods 0.000 claims description 7
- 238000000034 method Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 239000011521 glass Substances 0.000 claims description 2
- 239000000463 material Substances 0.000 claims description 2
- 238000012986 modification Methods 0.000 claims description 2
- 230000004048 modification Effects 0.000 claims description 2
- 239000010410 layer Substances 0.000 claims 18
- 238000005286 illumination Methods 0.000 claims 3
- 239000002356 single layer Substances 0.000 claims 1
- 230000001105 regulatory effect Effects 0.000 abstract description 4
- 230000010354 integration Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000001276 controlling effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005530 etching Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000010363 phase shift Effects 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 229910052710 silicon Inorganic materials 0.000 description 1
- 239000010703 silicon Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Landscapes
- Optical Integrated Circuits (AREA)
Abstract
The invention discloses a large-scale and reconfigurable three-dimensional integrated optical neural network, which belongs to the field of optical neural networks and comprises the following components: optical information input layer, reconfigurable hidden layer, detection output layer and feedback control structure: the reconfigurable hidden layer comprises a plurality of layers of cascaded waveguide structures, wherein each layer of waveguide structure comprises an independent modulation area and a continuous coupling area, and the optical waveguides of the independent modulation areas are independently transmitted and independently regulated by the optical modulator; after reaching the continuous coupling area, evanescent waves of the waveguide array are mutually and continuously coupled in the transmission direction and are integrally regulated and controlled by the optical modulator; the feedback control structure adjusts the loading signal of the optical modulator according to the detected spatial distribution information of the output light intensity so as to enable the reconfigurable hidden layer to identify the current input signal. The invention can greatly improve the actual computing power of the current on-chip integrated optical neural network, and simultaneously provides a solution to the problems of low reconfigurability and larger system size of the current three-dimensional space diffraction optical neural network.
Description
Technical Field
The invention belongs to the field of optical neural networks, and in particular relates to a large-scale reconfigurable three-dimensional integrated optical neural network.
Background
With the rapid increase of GPU computing speed and computing power, artificial intelligence has now become a central strength of a new technological revolution. However, the size and performance of electrical transistors have tended to be limited, and will not meet the ever-increasing data volume and power consumption limitations in reality in the future. Light has received great attention in the fields of communications and computing because of its wide bandwidth, high frequency, low energy consumption, and other advantages. Particularly, the characteristics of multidimensional resources and parallel transmission of light are widely applied to optical neural networks taking parallel matrix multiplication as a core, and the integrated optical neural networks are candidate methods for replacing GPU.
Currently, integrated optical neural networks on chip are mainly based on waveguide structure and tunable optical device integration schemes, such as Mach-Zehnder interferometer (Mach-Zehnder Interferometers, MZI) networks and micro-ring resonator (Microring resonators, MRR) arrays. This approach is highly integrated, but network complexity and total device count rise rapidly with increasing neuron population, with limited scalability, often limited to below 100 x 100.
Another approach is based on free-space optical diffraction, which uses each pixel of spatial light as an optical neuron. While the pixel count of a spatial light modulator can easily be 1000 x 1000, its integration still faces challenges such as the fact that a scheme that implements two-dimensional multi-plane diffraction by etching multiple rows of slit arrays on a silicon plate can only load and process one-dimensional information, greatly curtailing the large-scale advantage of the diffractive neural network, and reconstructing the effective phase shift of each slit is also difficult.
Disclosure of Invention
Aiming at the defects and improvement demands of the prior art, the invention provides a large-scale reconfigurable three-dimensional integrated optical neural network, which aims to fully utilize a spatial three-dimensional waveguide structure to realize large-scale neuron interconnection in the neural network and simultaneously realize the adjustment of the connection weight among neurons by applying an optical modulator, thereby constructing the large-scale reconfigurable three-dimensional integrated optical neural network and improving the expansibility and flexibility of the integrated optical neural network.
To achieve the above object, the present invention provides a large-scale reconfigurable three-dimensional integrated optical neural network, comprising: the device comprises an optical information input layer, a reconfigurable hidden layer, a detection output layer and a feedback control structure;
The reconfigurable hidden layer comprises m-level waveguide arrays corresponding to m full-connection layers in the optical neural network; each stage of the waveguide array comprises an independent modulation region and a continuous coupling region. The independent modulation area comprises N multiplied by M independent transmission waveguides and N multiplied by M optical modulators, the N multiplied by M independent transmission waveguides are used for transmitting input optical signals, no mutual coupling occurs between any two waveguides, the input optical signals are modulated by the N multiplied by M optical modulators, and the input optical signals correspond to reconfigurable (N multiplied by M) multiplied by (N multiplied by M) diagonal matrixes; the continuous coupling area comprises an N multiplied by M waveguide array, and the coupling relation between N multiplied by M inputs and N multiplied by M outputs of the waveguide array corresponds to a fixed N multiplied by M unitary matrix; the reconfigurable hidden layer is used for receiving the input optical signals of the optical signal input layer, and the intensity distribution of the input optical signals is regulated and controlled through the independent modulation area and the continuous coupling area in sequence to obtain output optical signals;
an optical information input layer for generating an input optical signal carrying input information and incident on the reconfigurable hidden layer;
And the detection output layer is used for detecting and outputting the optical signal.
And the feedback control structure is connected with the detection output layer and the reconfigurable hidden layer and is used for generating a feedback control signal according to the power distribution information of the output optical signal, adjusting the loading signal of the optical modulator by utilizing the feedback control signal, and further modulating the optical field signal in the N multiplied by M waveguide array of the multilayer cascade so as to train the reconfigurable hidden layer, wherein the reconfigurable hidden layer can identify the light intensity space distribution information of the current input signal. The continuous coupling region in the reconfigurable hidden layer can also be used to partially modulate the complex amplitude of the optical signal in the waveguide by the optical modulator, corresponding to the partially reconfigurable unitary matrix.
In one embodiment, the optical modulator may be a thermode, any other optical modulator (thermal modulation, electrical modulation), or a phase change material. For hot electrodes and other arbitrary optical modulators, the optical signal passing through the waveguide is modulated by applying other signals; for phase change materials, the transmission or reflection characteristics of the waveguide are adjusted by changing the phase state of the phase change material, so that effective control of the optical signal is realized.
In one embodiment, the optical information input layer includes: a light source, a spatial light modulator, a first beam transformer;
the light source is used for generating an original light spot;
The spatial light modulator is used for loading input vector information;
The first beam transformer is used for coupling the original light spot loaded with the input vector information into the reconfigurable hidden layer after being reduced.
In one embodiment, the probe output layer includes: a second beam transformer, a charge coupled device;
The beam converter is used for expanding the output optical signal;
The charge coupled device is used for detecting the power distribution of the output optical signal.
In one embodiment, the feedback control structure includes: a Field Programmable Gate Array (FPGA), a digital-to-analog/analog converter;
The Field Programmable Gate Array (FPGA) is used for executing a control algorithm;
The digital-analog/analog-digital converter is connected with the field programmable gate array FPGA and is used for converting digital level generated by the FPGA into analog level and controlling light field signals and input information in the waveguide in the reconfigurable hidden layer/converting the analog level of the detector array into digital level and then sending the digital level into the field programmable gate array FPGA for processing.
In general, through the above technical solutions conceived by the present invention, the following beneficial effects can be obtained:
(1) The reconfigurable three-dimensional integrated optical neural network based on the adjustable waveguide array comprises a multistage waveguide array with a reconfigurable hidden layer, and a plurality of full-connection layers corresponding to the neural network; each stage of the waveguide array comprises an independent modulation region and a continuous coupling region. An independent modulation region, wherein optical signals in the NxM independent transmission waveguides are modulated by the NxM optical modulators, and the independent modulation region corresponds to a reconfigurable (NxM) x (NxM) diagonal matrix; and a continuous coupling region, wherein the coupling relation between the N multiplied by M inputs and the N multiplied by M outputs of the waveguide array corresponds to a fixed (N multiplied by M) multiplied by (N multiplied by M) unitary matrix (which can also be modulated by an optical modulator so as to realize reconfiguration). And the feedback control structure generates a feedback control signal according to the power distribution of the output optical signal, and adjusts the loading signal of the optical modulator by using the feedback control signal, so as to control the optical modulator of the reconfigurable hidden layer, and further adjust the optical field signal in the waveguide, so as to train the reconfigurable hidden layer, and identify the current input signal. Therefore, the invention realizes the reconfigurable matrix multiplication calculation of the input information through the three-dimensional integrated adjustable waveguide array, further realizes the large-scale reconfigurable neural network, greatly improves the expansibility and flexibility of the integrated optical neural network, and simultaneously provides a solution for the problems of low reconfigurability and larger system size of the current three-dimensional space diffraction optical neural network.
(2) The large-scale reconfigurable three-dimensional integrated optical neural network chip provided by the invention realizes the improvement of actual computing power from an on-chip integrated square level to a three-dimensional integrated cubic level, and lays a foundation for realizing the large-scale integrated optical neural network.
(3) The large-scale reconfigurable three-dimensional integrated optical neural network chip provided by the invention is based on the architecture of the adjustable waveguide array, and can realize large-scale integration on a chip. The large-scale reconfigurable three-dimensional integrated optical neural network chip provided by the invention can be prepared in a glass-based material by adopting a femtosecond laser process.
Drawings
Fig. 1 is a schematic structural diagram of a large-scale reconfigurable three-dimensional integrated optical neural network chip provided in an example of the present invention, taking three layers 3*3 (n=3, m=3) as an example.
Fig. 2 is a side view of a structure of a large-scale reconfigurable three-dimensional integrated optical neural network chip provided in an example of the present invention, with three layers 3*3 (n=3, m=3) as an example.
Fig. 3 is a top view of a structure of a large-scale reconfigurable three-dimensional integrated optical neural network chip provided in an example of the present invention, taking three layers 3*3 (n=3, m=3) as an example.
Fig. 4 is a schematic diagram of an experimental setup of a large-scale reconfigurable three-dimensional integrated optical neural network chip provided in an example of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. In addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not interfere with each other.
The invention relates to a large-scale reconfigurable three-dimensional integrated optical neural network chip, which comprises:
an optical information input layer I for generating an input optical signal carrying input information and incident on the reconfigurable hidden layer II;
And the reconfigurable hidden layer II is used for regulating and controlling the intensity distribution of the input optical signal by utilizing the waveguide array therein. The optical neural network comprises an M-level waveguide array, corresponding to M full connection layers in the optical neural network; each layer of waveguide array comprises an independent modulation region 4 and a continuous coupling region 5. An independent modulation area 4, wherein optical signals in the NxM independent transmission waveguides are modulated by the NxM optical modulators, and the optical signals correspond to reconfigurable (NxM) x (NxM) diagonal matrixes; a continuous coupling region 5 for coupling the n×m inputs and n×m outputs of the waveguide array, corresponding to a fixed (n×m) × (n×m) unitary matrix;
and the detection output layer III is used for detecting and outputting the optical signal.
And the feedback control structure IV is connected with the detection output layer III and the reconfigurable hidden layer II and is used for generating a feedback control signal according to the power distribution information of the output optical signal to be detected, adjusting the loading signal of the optical modulator by utilizing the feedback control signal and further modulating the optical field signal in the multi-layer cascaded waveguide array so as to train the reconfigurable hidden layer II, and the reconfigurable hidden layer II can identify the light intensity space distribution information of the current input signal.
The structure shown in fig. 1 depicts only a three-layer cascaded 3 x 3 (n=3, m=3) waveguide array. The number of cascade layers and the number of waveguides can be expanded according to the scheme.
Fig. 2 and 3 are side and top views, respectively, of the schematic structure of fig. 1.
In one embodiment, the optical modulator may be a thermode, any other optical modulator (thermal modulation, electrical modulation), or a phase change material. For hot electrodes and other arbitrary optical modulators, the optical signal passing through the waveguide is modulated by applying other signals; for phase change materials, the transmission or reflection characteristics of the waveguide are adjusted by changing the phase state of the phase change material, so that effective control of the optical signal is realized.
In particular, assuming that the number of waveguide arrays of the reconfigurable hidden layer II is nxm, the spatial light modulator 2 may divide the spatial distribution of the input light into nxm pixels, each represented by the complex amplitude of the light at this location, then the input may be written as a vector E in of (nxm) x 1, the transmission of the light within the independent modulation region 4 may be described by a diagonal matrix M i (i=1, 2 …) of reconfigurable (nxm) x (nxm), the transmission of the light within the continuous coupling region 5 may be regarded as a unitary matrix U i (i=1, 2 …) of fixed (nxm) x (nxm), and the effect of the photo-coupling element 7 is to take the intensity I out of the output light signal, so that the information received by the M-layer three-dimensional integrated optical neural network detector may be represented as :Iout=|MmUmMm-1…M2U2M1U1Ein|2, the whole system may be regarded as first multiplying the input vector by a matrix and then producing a square non-linear, which may also be regarded as a part of the neural network, and thus may be fully composed of a neural network.
In one embodiment, the optical information input layer I includes:
The spatial light modulator 2 performs loading of the initial input signal by encoding the phase of the input light source and transfers it to the spatial distribution of the input light;
The beam transformer 3 is disposed behind the spatial light modulator 2, and is configured to adjust the spot size of the light beam after loading the input signal, so as to match the size of the adjustable waveguide array of the reconfigurable hidden layer II, thereby realizing efficient coupling input.
In one embodiment, the detection output layer III includes:
the beam converter 6 expands the spot size of the beam of the output signal of the reconfigurable hidden layer II, so that the next detection is facilitated;
the charge coupled device 7 is used to detect the power distribution of the output optical signal.
In one embodiment, a feedback control structure IV is connected to the detection output layer III and the reconfigurable hidden layer II, and is configured to generate a feedback control signal according to the power distribution information of the optical signal output by the waveguide to be tested, and adjust the loading signal of the optical modulator by using the feedback control signal. The feedback control structure IV includes:
The Field Programmable Gate Array (FPGA) is used for executing a control algorithm;
The digital-analog/analog-digital converter is connected with the FPGA and is used for converting the digital level generated by the FPGA into the analog level and controlling the light field signal and the input information in the waveguide in the reconfigurable hidden layer II/converting the analog level of the detector array into the digital level and then sending the digital level into the FPGA for processing.
The experimental setup diagram shown in fig. 4 contains the various parts of the setup mentioned in the above embodiments.
The whole three-dimensional integrated optical neural network provided by the invention can work in two modes: a pre-training mode and an online training mode. The pre-training mode needs to obtain the waveguide array coupling matrix M and the modulation effect of the light modulator on light from simulation or experiments in advance, a simulation model is built, and the signal of each light modulator is optimized by adopting a gradient descent algorithm in a computer so as to realize target output. And then the optimized signals are configured on the optical modulator, and the construction of the optical neural network is completed. The on-line training mode refers to that the actual physical system transmission is directly used for replacing the mathematical model transmission of the method, so that the method has stronger robustness to the defects of the physical system.
In summary, the technical point of the application is that the design of the reconfigurable hidden layer can realize random distribution of the propagation path of the chip input optical signal, the high density, high flexibility and expandable characteristic of the three-dimensional integrated optical waveguide determine the potentially extremely strong light field regulation capability, and the reconfigurable function of the optical neural network chip can be completed by establishing the mapping relation between the abundant target functional characteristic and the reconfigurable modulation physical characteristic of the neural network, thereby being beneficial to accelerating the realization of the optical neural network system with high calculation power, high integration level and multiple functions.
The waveguide array and one of the optical modulator types, the thermode, of the present invention can be fabricated using a femtosecond laser machining process.
It will be readily appreciated by those skilled in the art that the foregoing description is merely a preferred embodiment of the invention and is not intended to limit the invention, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (10)
1. A large-scale reconfigurable three-dimensional integrated optical neural network, comprising: an optical information input layer (I), a reconfigurable hidden layer (II), a detection output layer (III) and a feedback control structure (IV);
The reconfigurable hidden layer (II) comprises a multi-layered cascaded waveguide array; each layer of cascaded waveguide array comprises an independent modulation area and a continuous coupling area, wherein the independent modulation area comprises N multiplied by M independent transmission waveguides and N multiplied by M optical modulators, the N multiplied by M independent transmission waveguides are used for transmitting input optical signals, mutual coupling does not occur between any two waveguides, the optical modulators which correspond to one independently modulate the transmission light in each independent transmission waveguide, and the optical modulators correspond to reconfigurable (N multiplied by M) multiplied by (N multiplied by M) diagonal matrixes; the continuous coupling area comprises a waveguide array formed by N multiplied by M waveguides, and the waveguide array corresponds to a fixed (N multiplied by M) multiplied by (N multiplied by M) unitary matrix; the input optical signals are modulated alternately through a single-layer neural network architecture formed by cascading the independent modulation areas and the continuous coupling areas, so that output optical signals to be detected are obtained;
The optical information input layer (I) is used for generating an input optical signal carrying input image information and being incident at normal incidence on the reconfigurable hidden layer (II);
the detection output layer (III) is used for detecting the power of the output optical signal to be detected;
The feedback control structure (IV) is connected with the detection output layer (III) and the reconfigurable hidden layer (II) and is used for generating a feedback control signal according to the power distribution information of the output optical signal to be detected, adjusting the loading signal of the optical modulator by utilizing the feedback control signal, and further modulating the optical field signal in the multi-layer cascade waveguide array, so that the reconfigurable hidden layer (II) can perform reconfigurable modulation on the input light to obtain the output optical signal to be detected.
2. The large scale reconfigurable three-dimensional integrated optical neural network of claim 1, wherein the three-dimensional integrated optical neural network is implemented by writing optical waveguides in a glass-based material by a femtosecond laser modification process.
3. A large-scale reconfigurable three-dimensional integrated optical neural network according to claim 1, wherein the optical information input layer (I) comprises: an illumination light source (1), a spatial light modulator (2), and a beam transformer (3);
The illumination source (1) is used for generating an original light spot;
The spatial light modulator (2) is used for loading input vector information;
The beam transformer (3) is used for shrinking an original light spot of the loaded information and then coupling the reduced original light spot into the reconfigurable hidden layer (II);
When the device works, an illumination light source (1) transmits input vector information loaded by a spatial light modulator (2), and then light spots loaded with the information are reduced by a beam converter (3), so that a reconfigurable hidden layer (II) formed by a beam coupling input waveguide array containing the input vector information is realized.
4. A large-scale reconfigurable three-dimensional integrated optical neural network according to any of claims 1-3, characterized in that the detection output layer (III) comprises a beam transformer (6) and a charge-coupled element (7), and that light after output from the reconfigurable hidden layer (II) passes through the beam transformer (6) to expand the original output spot to a complete coverage of the charge-coupled element (7).
5. A large-scale reconfigurable three-dimensional integrated optical neural network according to any of claims 1-3, characterized in that the independent modulation zone (4) comprises N x M waveguides independently transmitting independent modulation, corresponding to N x M neurons, N x M optical modulators modulating the complex amplitude of the optical signal in N x M waveguides in a one-to-one correspondence, corresponding to a reconfigurable (N x M) x (N x M) diagonal matrix.
6. A large scale reconfigurable three-dimensional integrated optical neural network according to any of claims 1-3, characterized in that the continuous coupling region (5) comprises N x M waveguides, corresponding to N x M neurons, which are integrally modulated by mutual coupling, the array of continuously coupled waveguides reaching a certain transmission length, enabling the coupling between the light field of any input waveguide to any output waveguide, corresponding to a fixed (N x M) x (N x M) unitary matrix.
7. A large scale reconfigurable three-dimensional integrated optical neural network according to any of claims 1 to 6, comprising:
Utilizing N×M optical modulators of an independent modulation region (4) to regulate and control optical field signals in N×M waveguides in one-to-one correspondence, and corresponding to a reconfigurable (N×M) x (N×M) diagonal matrix; according to the waveguide array arrangement mode and the interval distance of the continuous coupling area (5), the coupling relation between the N multiplied by M inputs and the N multiplied by M outputs of the waveguide array is calculated, and a fixed (N multiplied by M) multiplied by (N multiplied by M) unitary matrix is correspondingly obtained;
The signals loaded on the optical modulators of the independent modulation areas (4) are controlled by a feedback control structure (IV) to achieve the adjustment of the reconfigurable (nxm) × (nxm) diagonal matrix.
8. The large-scale reconfigurable three-dimensional integrated optical neural network of claim 7, wherein the feedback control structure (IV) comprises:
The Field Programmable Gate Array (FPGA) is used for executing a control algorithm;
The digital-analog/analog-digital converter is connected with the FPGA and is used for converting the digital level generated by the FPGA into an analog level and controlling the optical modulator signal/converting the analog level of the detection output layer (III) into the digital level and then sending the digital level into the FPGA for processing.
9. The large scale reconfigurable three-dimensional integrated optical neural network of claim 1, wherein the optical modulator is a thermode, a thermo-optic modulator, an electro-optic modulator, or a phase change material, for which the optical signal passing through the waveguide is modulated by applying other signals; for phase change materials, the transmission or reflection characteristics of the waveguide are adjusted by changing the phase state of the phase change material, so that effective control of the optical signal is realized.
10. A large scale reconfigurable three-dimensional integrated optical neural network according to any of claims 1 to 6, wherein the nxm waveguide arrays are uniformly arranged in square grid arrays or distributed in circular, polygonal grid arrays.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410284681.2A CN118246503A (en) | 2024-03-13 | 2024-03-13 | Large-scale reconfigurable three-dimensional integrated optical neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410284681.2A CN118246503A (en) | 2024-03-13 | 2024-03-13 | Large-scale reconfigurable three-dimensional integrated optical neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN118246503A true CN118246503A (en) | 2024-06-25 |
Family
ID=91553352
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410284681.2A Pending CN118246503A (en) | 2024-03-13 | 2024-03-13 | Large-scale reconfigurable three-dimensional integrated optical neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN118246503A (en) |
-
2024
- 2024-03-13 CN CN202410284681.2A patent/CN118246503A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bai et al. | Photonic multiplexing techniques for neuromorphic computing | |
WO2021227767A1 (en) | All-optical diffractive neural network and system implemented on optical waveguide and/or optical chip | |
KR100871520B1 (en) | Composite material with controllable resonant cells | |
CN112506265A (en) | Light calculation device and calculation method | |
CN112001487A (en) | Photon neural network | |
WO2023201970A1 (en) | Computing chip, system, and data processing method | |
JP2024503991A (en) | Balanced photonic architecture for matrix computation | |
CN112101540A (en) | Optical neural network chip and calculation method thereof | |
Liu et al. | Silicon optical phased array side lobe suppression based on an improved genetic algorithm | |
Nikkhah et al. | Reconfigurable nonlinear optical element using tunable couplers and inverse-designed structure | |
US20240078419A1 (en) | Optical neuron unit and network of the same | |
CN211880393U (en) | Planar sparse light-controlled phased array transmitting antenna system with low array element number | |
CN118246503A (en) | Large-scale reconfigurable three-dimensional integrated optical neural network | |
WO2024021659A1 (en) | Three-dimensional photonic chip architecture based on vcsel array, application, and method for calculating structure of dnns | |
TWI832681B (en) | Photonic computing system | |
CN116739063A (en) | Neural network accelerator based on multimode interferometer and coherent detection | |
US20230163859A1 (en) | Optical routing network-based quantum array control | |
Chen et al. | Hybrid optical-electronic neural network with pseudoinverse learning for classification inference | |
Ma et al. | Intelligent neuromorphic computing based on nanophotonics and metamaterials | |
Hermans et al. | Towards trainable media: Using waves for neural network-style training | |
US11567177B2 (en) | Optical phased array lidar | |
CN112232487A (en) | Optical neural network chip and calculation method thereof | |
US11836609B2 (en) | Physical neural network of optical resonators and waveguides | |
Ma et al. | Deep learning in metasurface design and optimization | |
WO2023039858A1 (en) | Optical computing system, optical computing method and control apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination |