CN113610224B - Method, system, equipment and medium for maximizing pooling of optical neural network - Google Patents
Method, system, equipment and medium for maximizing pooling of optical neural network Download PDFInfo
- Publication number
- CN113610224B CN113610224B CN202110808862.7A CN202110808862A CN113610224B CN 113610224 B CN113610224 B CN 113610224B CN 202110808862 A CN202110808862 A CN 202110808862A CN 113610224 B CN113610224 B CN 113610224B
- Authority
- CN
- China
- Prior art keywords
- uploading
- type micro
- ring
- downloading type
- downloading
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/067—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means
- G06N3/0675—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means using electro-optical, acousto-optical or opto-electronic means
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a method, a system, equipment and a storage medium for maximizing pooling of an optical neural network, wherein the method comprises the following steps: determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to realize the maximum pooling. The invention realizes the maximum pooling by utilizing the MRR of the Add-Drop type, provides an optical solution of a key module for all-optical artificial intelligent computation, and has the advantages of high speed and low power consumption.
Description
Technical Field
The present invention relates to the field of optical neural networks, and more particularly, to a method, a system, a computer device, and a readable medium for maximizing pooling of an optical neural network.
Background
With the development of technology, the society now enters the era of cloud+ai+5g, and in order to realize the operation requirement of cloud+ai+5g, a special chip supporting a large amount of operation is required. The development of the semiconductor lithography process level is the basic stone of an electronic computer with a chip as a core, the current manufacturing process of semiconductor lithography is almost the physical limit of moore's law, and along with the smaller and smaller manufacturing process, the transistor units in the chip are approaching to the molecular scale, and the bottleneck effect of the semiconductor manufacturing process is more and more obvious.
With globalization and high-speed development of technology, the amount of data to be processed is rapidly increasing, and corresponding data processing models and algorithms are also increasing, so that the demands on computing power and power consumption are continuously increasing. However, the existing von neumann architecture and harvard architecture electronic computers have problems of transmission bottleneck, power consumption increase, calculation power bottleneck and the like, so that it is more and more difficult to meet the demands of large data time for calculation power and power consumption, and therefore, the improvement of the operation speed and the reduction of the operation power consumption are critical problems at present.
Optical computing has many advantages over electrical computing, such as: the optical signal is transmitted at the speed of light, so that the speed is greatly improved; the light has natural parallel processing capability and mature wavelength division multiplexing technology, so that the data processing capability, capacity and bandwidth are greatly improved; the optical calculation power consumption is hopefully 10-18J/bit, and the photon device is hundreds of times faster than the electronic device under the same power consumption. The photon computing chip has the advantages of high-speed parallelism and low power consumption by taking photons as carriers of information, so that the photon computing chip is considered as the most promising scheme for high-speed, large-data-volume and artificial intelligent computing processing in the future. However, no method for solving the maximum pooling by utilizing the optical signal is currently known.
Disclosure of Invention
In view of the above, an object of the embodiments of the present invention is to provide a method, a system, a computer device and a computer readable storage medium for maximizing pooling of an optical neural network, which uses an optical intensity transfer function of an MRR as a basis for implementing the optical neural network, changes an effective refractive index of the MRR by current to further change the MRR transfer function, implements an analog solution applicable to maximizing pooling in the optical neural network, implements maximum pooling by using Add-Drop MRR, provides an optical solution of a key module for all-optical artificial intelligent computation, and has advantages of high speed and low power consumption.
Based on the above objects, an aspect of the embodiments of the present invention provides a method for maximizing pooling of an optical neural network, including the steps of: determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to realize the maximum pooling.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
In another aspect of the embodiments of the present invention, there is provided a system for maximizing pooling of an optical neural network, including: the setting module is configured to determine the number of uploading/downloading type micro-rings according to the pooling size and adjust the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; the Input module is configured to Input corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and Input optical signals output by the uploading/downloading type micro-rings into the photodiodes to obtain corresponding output currents; and the adjusting module is configured to determine the largest output current in the output currents and adjust the phase of the uploading/downloading type micro-ring corresponding to the largest output current so as to realize the maximum pooling.
In some embodiments, the adjustment module is configured to: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjustment module is configured to: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, the setup module is configured to: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
In yet another aspect of the embodiment of the present invention, there is also provided a computer apparatus, including: at least one processor; and a memory storing computer instructions executable on the processor, which when executed by the processor, perform the steps of the method as above.
In yet another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium storing a computer program which, when executed by a processor, implements the method steps as described above.
The invention has the following beneficial technical effects: the optical neural network is based on the light intensity transfer function of the MRR, the effective refractive index of the MRR is changed through current, the MRR transfer function is changed, the simulation solution suitable for the maximum pooling in the optical neural network is realized, the maximum pooling is realized by utilizing the Add-Drop type MRR, the optical solution of a key module is provided for all-optical artificial intelligent computation, and the optical neural network has the advantages of high speed and low power consumption.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are necessary for the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention and that other embodiments may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of an embodiment of a method for maximizing pooling of an optical neural network provided by the present invention;
FIG. 2 is a schematic diagram of maximum pooling;
FIG. 3 is a schematic diagram of an upload/download type (Add-Drop type) micro-ring;
FIG. 4 is a schematic diagram showing the light intensity distribution of the uploading/downloading type micro-ring under the non-resonance (a) and resonance (b) conditions;
FIG. 5a shows an upload/download micro-ring transfer function T p With phase
Is a variation of the schematic diagram;
FIG. 5b shows an upload/download micro-ring transfer function T d With phase
Is a variation of the schematic diagram;
fig. 6 is a schematic diagram of an architecture for implementing 2 x 2 max pooling of upload/download micro-loops;
FIG. 7 is a schematic diagram of a hardware architecture of an embodiment of a computer device for maximizing pooling of an optical neural network according to the present invention;
fig. 8 is a schematic diagram of an embodiment of a computer storage medium for maximizing pooling of an optical neural network provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the following embodiments of the present invention will be described in further detail with reference to the accompanying drawings.
It should be noted that, in the embodiments of the present invention, all the expressions "first" and "second" are used to distinguish two entities with the same name but different entities or different parameters, and it is noted that the "first" and "second" are only used for convenience of expression, and should not be construed as limiting the embodiments of the present invention, and the following embodiments are not described one by one.
In a first aspect of the embodiments of the present invention, an embodiment of a method for maximizing pooling of an optical neural network is provided. Fig. 1 is a schematic diagram of an embodiment of a method for maximizing pooling of an optical neural network provided by the present invention. As shown in fig. 1, the embodiment of the present invention includes the following steps:
s1, determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light;
s2, inputting corresponding Input light to an Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and
s3, determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to achieve maximum pooling.
Pooling is an important operator in neural networks, the pooling can reduce the size of the feature map and keep some invariance, such as rotation, translation, and expansion, common pooling methods include maximum pooling and average pooling, where maximum pooling calculates the maximum value of neurons in the pooled kernel as output, fig. 2 shows a schematic diagram of maximum pooling, and the pooling size in fig. 2 is 2×2.
One operation in maximum pooling can be expressed as
Where n is the size of the pooling nucleus, x i For the pooling layer input neurons, y is the pooling layer output neurons, when n=2, one operation in the pooling layer can be expressed as: y=max (x 1 ,x 2 ,x 3 ,x 4 )。
Fig. 3 is a schematic diagram of an upload/download type (Add-Drop type) micro-ring, and as shown in fig. 3, the upload/download type micro-ring is composed of one micro-ring waveguide and two linear access waves, and has four ports in total: input port, thru (via) port, add (upload) port, and Drop (download) port. The present invention assumes that the Add port has no optical signal.
Fig. 4 is a schematic diagram showing light intensity distribution in the case of uploading/downloading micro-ring non-resonance (a) and resonance (b). As shown in fig. 4, an Input optical signal enters from the Input port, and when the wavelength of the incident light satisfies the resonance condition, most of the optical signal of the wavelength is output from the Drop port, and almost no output is output from the thr port; if the wavelength of the incident light does not meet the resonance condition, only a weak optical field enters the micro-ring cavity, most of the input light is output from the Threu port, and the Drop port has only a small light output.
When light is transmitted in the micro-ring, the light is limited by the micro-ring, and the optical path difference generated when the light is transmitted around the micro-ring is as followsResonance occurs when the wavelength of the optical signal is an integer multiple of the wavelength of the optical signal, the intensity of the optical signal is continuously enhanced, and the condition that the optical signal is enhanced by interaction is called a resonance condition, and the resonance equation of the micro-ring is as follows: 2 pi Rn eff ＝mλ i . Lambda in i R is the radius of the uploading/downloading type micro-ring, n is the wavelength eff And m is a positive integer, which is the effective refractive index of light. Light of a wavelength satisfying the above formula, i.e., satisfying the resonance condition, is confined in the micro-ring. When current is passed through the uploading/downloading type micro-ring, the uploading/downloading type micro-ring is heated, resulting in effective refractive index n of light eff Changes occur, thereby shifting the resonant wavelength.
The expression of the transfer function of the intensity of light passing through the Thru port and the intensity of light entering the Input port of the Add-Drop type MRR is as follows:
the transfer function of the intensity of light passing through the Drop port and the intensity of light entering the Input port of an Add-Drop type MRR is expressed as follows:
wherein phi is i For the phase of the MRR, r is the self-coupling coefficient, a defines the propagation loss of the loop and the directional coupler. In the case of negligible coupling loss, i.e. a.apprxeq.1, the relation between the transfer function of the light of the Thu port and the Drop transfer function of the light of the Drop port is T p ＝1-T q 。
FIG. 5a shows an upload/download micro-ring transfer function T p With phase
Is a variation of the schematic diagram; FIG. 5b shows the upload/download micro-ring transfer function T d With phase->
Is a variation of the schematic diagram. As shown in fig. 5a and 5b, the upload/download micro-ring transfer function T p ,T d Along with the phase phi i Is varied at [0,1 ]]And changes between. When current is passed through the uploading/downloading type micro-ring, the uploading/downloading type micro-ring is heated, resulting in n eff Thereby resulting in a phase phi i Is a function of the transfer function T which ultimately affects the light intensity p ,T d 。
Fig. 6 shows a schematic diagram of an architecture for implementing 2×2 max pooling of upload/download micro-loops. An embodiment of the present invention will be described with reference to fig. 6.
The number of uploading/downloading type micro-rings is determined according to the pooling size, and the radius of the uploading/downloading type micro-rings is adjusted according to the wavelength of each input light. In the embodiment of the present invention, taking 2×2 as an example, but not limiting the invention, the number of uploading/downloading type micro-rings may be determined according to the pooling size, for example, the pooling size is 2×2, the number of uploading/downloading type micro-rings may be 4, the pooling size is 3*3, the number of uploading/downloading type micro-rings may be 9, the pooling size is 4*4, the number of uploading/downloading type micro-rings may be 16, and so on.
In some embodiments, said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light. When the wavelength lambda of the input optical signal i Meets 2 pi Rn between the ring radius R in the uploading/downloading type micro-ring eff ＝mλ i Resonance may be generated at this time, and thus, the radius of the uploading/downloading type micro-ring may be adjusted according to the wavelength of each input light.
And inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into the photodiodes to obtain corresponding output currents. Four input light intensities |E 1 | 2 ,|E 2 | 2 ,|E 3 | 2 ,|E 4 | 2 Respectively at lambda 1 ,λ 2 ,λ 3 ,λ 4 The wavelength of (2) is input at input end, and respectively resonates with four uploading/downloading micro-rings, and the optical signal is transmitted from Drop port, at this time, the optical signal is fed into a photodiode with spectral response of Re, and then the induced current of the photodiode is as follows:
and determining the maximum output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the maximum output current to realize maximum pooling.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0. In the control module, the output current is compared, and the phase phi of the uploading/downloading micro-ring with the largest induced current is adjusted i Let the transfer function T p ＝1,T d =0, only the optical signal with the largest induced current (the largest light intensity) is output at the Thru end, and 2×2 max pooling is achieved.
The embodiment of the invention takes the light intensity transfer function of the MRR as the basis for realizing the optical neural network, changes the effective refractive index of the MRR through current so as to change the transfer function of the MRR, realizes an analog solution suitable for the maximum pooling in the optical neural network, realizes the maximum pooling by utilizing the Add-Drop type MRR, provides an optical solution of a key module for all-optical artificial intelligent computation, and has the advantages of high speed and low power consumption.
It should be noted that, in the embodiments of the method for maximizing pooling of an optical neural network, the steps may be intersected, replaced, added and deleted, so that the method for maximizing pooling of an optical neural network by using these reasonable permutation and combination transforms shall also belong to the protection scope of the present invention, and shall not limit the protection scope of the present invention to the embodiments.
Based on the above object, a second aspect of the embodiments of the present invention provides a system for maximizing pooling of an optical neural network, including: the setting module is configured to determine the number of uploading/downloading type micro-rings according to the pooling size and adjust the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; the Input module is configured to Input corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and Input optical signals output by the uploading/downloading type micro-rings into the photodiodes to obtain corresponding output currents; and the adjusting module is configured to determine the largest output current in the output currents and adjust the phase of the uploading/downloading type micro-ring corresponding to the largest output current so as to realize the maximum pooling.
In some embodiments, the adjustment module is configured to: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjustment module is configured to: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, the setup module is configured to: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
The embodiment of the invention takes the light intensity transfer function of the MRR as the basis for realizing the optical neural network, changes the effective refractive index of the MRR through current so as to change the transfer function of the MRR, realizes an analog solution suitable for the maximum pooling in the optical neural network, realizes the maximum pooling by utilizing the Add-Drop type MRR, provides an optical solution of a key module for all-optical artificial intelligent computation, and has the advantages of high speed and low power consumption.
In view of the above object, a third aspect of the embodiments of the present invention provides a computer device, including: at least one processor; and a memory storing computer instructions executable on the processor, the instructions being executable by the processor to perform the steps of: s1, determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; s2, inputting corresponding Input light to an Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and S3, determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to realize the maximum pooling.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
The embodiment of the invention takes the light intensity transfer function of the MRR as the basis for realizing the optical neural network, changes the effective refractive index of the MRR through current so as to change the transfer function of the MRR, realizes an analog solution suitable for the maximum pooling in the optical neural network, realizes the maximum pooling by utilizing the Add-Drop type MRR, provides an optical solution of a key module for all-optical artificial intelligent computation, and has the advantages of high speed and low power consumption.
As shown in fig. 2, a hardware structure diagram of an embodiment of the optical neural network maximizing pooling computer device provided by the present invention is shown.
Taking the example of the apparatus shown in fig. 2, the apparatus includes a processor 201 and a memory 202, and may further include: an input device 203 and an output device 204.
The processor 201, memory 202, input devices 203, and output devices 204 may be connected by a bus or other means, for example in fig. 2.
The memory 202 is used as a non-volatile computer readable storage medium for storing non-volatile software programs, non-volatile computer executable programs, and modules, such as program instructions/modules corresponding to the method for maximizing pooling of optical neural networks in the embodiments of the present application. The processor 201 executes various functional applications of the server and data processing, i.e., a method of achieving the maximization of the optical neural network of the above-described method embodiment, by running nonvolatile software programs, instructions, and modules stored in the memory 202.
The input device 203 may receive input information such as a user name and a password. The output device 204 may include a display device such as a display screen.
Program instructions/modules corresponding to one or more methods of maximizing pooling of optical neural networks are stored in memory 202, which when executed by processor 201, perform a method of maximizing pooling of optical neural networks, the method comprising the steps of: determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to realize the maximum pooling.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
Any one embodiment of a computer device that performs the method of maximizing pooling of an optical neural network described above may achieve the same or similar effects as any of the method embodiments described above that correspond thereto.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor performs a method of maximising pooling of an optical neural network. The method comprises the following steps: determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light; inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and determining the largest output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the largest output current to realize the maximum pooling.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
In some embodiments, the adjusting the phase of the uploading/downloading micro-ring corresponding to the maximum output current to achieve the maximum pooling includes: and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
In some embodiments, said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises: the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
As shown in fig. 3, a schematic diagram of an embodiment of the optical neural network maximizing pooling computer storage medium is provided in the present invention. Taking a computer storage medium as shown in fig. 3 as an example, the computer readable storage medium 3 stores a computer program 31 that, when executed by a processor, performs the method as described above.
Finally, it should be noted that, as will be understood by those skilled in the art, all or part of the processes in the methods of the embodiments described above may be implemented by a computer program to instruct related hardware, and the program of the method for maximizing pooling of the optical neural network may be stored in a computer readable storage medium, where the program may include processes in the embodiments of the methods described above when executed. The storage medium of the program may be a magnetic disk, an optical disk, a read-only memory (ROM), a random-access memory (RAM), or the like. The computer program embodiments described above may achieve the same or similar effects as any of the method embodiments described above.
The foregoing is an exemplary embodiment of the present disclosure, but it should be noted that various changes and modifications could be made herein without departing from the scope of the disclosure as defined by the appended claims. The functions, steps and/or actions of the method claims in accordance with the disclosed embodiments described herein need not be performed in any particular order. Furthermore, although elements of the disclosed embodiments may be described or claimed in the singular, the plural is contemplated unless limitation to the singular is explicitly stated.
It should be understood that as used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly supports the exception. It should also be understood that "and/or" as used herein is meant to include any and all possible combinations of one or more of the associated listed items.
The foregoing embodiment of the present invention has been disclosed with reference to the number of embodiments for the purpose of description only, and does not represent the advantages or disadvantages of the embodiments.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, and the program may be stored in a computer readable storage medium, where the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
Those of ordinary skill in the art will appreciate that: the above discussion of any embodiment is merely exemplary and is not intended to imply that the scope of the disclosure of embodiments of the invention, including the claims, is limited to such examples; combinations of features of the above embodiments or in different embodiments are also possible within the idea of an embodiment of the invention, and many other variations of the different aspects of the embodiments of the invention as described above exist, which are not provided in detail for the sake of brevity. Therefore, any omission, modification, equivalent replacement, improvement, etc. of the embodiments should be included in the protection scope of the embodiments of the present invention.
Claims (10)
1. A method for maximizing pooling of an optical neural network, comprising the steps of:
determining the number of uploading/downloading type micro-rings according to the pooling size, and adjusting the radius of the uploading/downloading type micro-rings according to the wavelength of each input light;
inputting corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and inputting optical signals output by the uploading/downloading type micro-rings into a photodiode to obtain corresponding output currents; and
and determining the maximum output current in the output currents, and adjusting the phase of the uploading/downloading type micro-ring corresponding to the maximum output current to realize maximum pooling.
2. The method of claim 1, wherein adjusting the phase of the upload/download micro-loop corresponding to the maximum output current to achieve maximum pooling comprises:
and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
3. The method of claim 2, wherein adjusting the phase of the upload/download micro-loop corresponding to the maximum output current to achieve maximum pooling comprises:
and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
4. The method of claim 1, wherein said adjusting the radius of the upload/download micro-ring according to the wavelength of each input light comprises:
the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
5. A system for maximizing pooling of an optical neural network, comprising:
the setting module is configured to determine the number of uploading/downloading type micro-rings according to the pooling size and adjust the radius of the uploading/downloading type micro-rings according to the wavelength of each input light;
the Input module is configured to Input corresponding Input light to the Input end of each uploading/downloading type micro-ring respectively, and Input optical signals output by the uploading/downloading type micro-rings into the photodiodes to obtain corresponding output currents; and
and the adjusting module is configured to determine the largest output current in the output currents and adjust the phase of the uploading/downloading type micro-ring corresponding to the largest output current so as to realize the maximum pooling.
6. The system of claim 5, wherein the adjustment module is configured to:
and current is introduced into the uploading/downloading type micro-ring so as to adjust the phase of the uploading/downloading type micro-ring.
7. The system of claim 6, wherein the adjustment module is configured to:
and adjusting the phase of the uploading/downloading type micro-ring so that the value of the transfer function of the light intensity passing through the Thu end and the light intensity entering the Input end is 1, and the value of the transfer function of the light intensity passing through the Drop end and the light intensity entering the Input end is 0.
8. The system of claim 5, wherein the setup module is configured to:
the radius of the uploading/downloading type micro-ring is adjusted so that the product of the circumference of the uploading/downloading type micro-ring and the effective refractive index of the light is an integral multiple of the wavelength of the input light.
9. A computer device, comprising:
at least one processor; and
a memory storing computer instructions executable on the processor, which when executed by the processor, perform the steps of the method of any one of claims 1-4.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110808862.7A CN113610224B (en) | 2021-07-16 | 2021-07-16 | Method, system, equipment and medium for maximizing pooling of optical neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110808862.7A CN113610224B (en) | 2021-07-16 | 2021-07-16 | Method, system, equipment and medium for maximizing pooling of optical neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113610224A CN113610224A (en) | 2021-11-05 |
| CN113610224B true CN113610224B (en) | 2023-05-23 |
Family
ID=78304772
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110808862.7A Active CN113610224B (en) | 2021-07-16 | 2021-07-16 | Method, system, equipment and medium for maximizing pooling of optical neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113610224B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635941A (en) * | 2018-11-06 | 2019-04-16 | 华中科技大学 | A kind of maximum pond processor based on 1T1R memory device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112825259A (en) * | 2019-11-20 | 2021-05-21 | 三星电子株式会社 | Nonvolatile memory and operating method thereof |
-
2021
- 2021-07-16 CN CN202110808862.7A patent/CN113610224B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635941A (en) * | 2018-11-06 | 2019-04-16 | 华中科技大学 | A kind of maximum pond processor based on 1T1R memory device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113610224A (en) | 2021-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Photonic matrix multiplication lights up photonic accelerator and beyond | |
| Wu et al. | Programmable phase-change metasurfaces on waveguides for multimode photonic convolutional neural network | |
| Rakshit et al. | Design of ring resonator based all optical switch for logic and arithmetic operations–a theoretical study | |
| Zhou et al. | Chip-scale optical matrix computation for PageRank algorithm | |
| CN109639359A (en) | Photon neural network convolutional layer chip based on micro-ring resonator | |
| Yan et al. | All-optical graph representation learning using integrated diffractive photonic computing units | |
| CN102023672B (en) | Integrated optical vector-matrix multiplier based on micro-ring resonators | |
| Nahmias et al. | Neuromorphic photonics | |
| CN113961035A (en) | Deep learning-oriented high-efficiency reconfigurable all-optical neural network computing chip architecture | |
| Ben-Ali et al. | Accordable filters by defect modes in single and double negative star waveguides grafted dedicated to electromagnetic communications applications | |
| Zhou et al. | Tunable on-chip mode converter enabled by inverse design | |
| Geng et al. | N-port strictly non-blocking optical router based on Mach-Zehnder optical switch for photonic networks-on-chip | |
| CN113610224B (en) | Method, system, equipment and medium for maximizing pooling of optical neural network | |
| Li et al. | An energy-efficient silicon photonic-assisted deep learning accelerator for big data | |
| 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 | |
| Feng et al. | Integrated multi-operand optical neurons for scalable and hardware-efficient deep learning | |
| CN113392966B (en) | Method, equipment and storage medium for realizing neural network average pooling | |
| Dehghani et al. | ICES: an innovative crosstalk-efficient 2× 2 photonic-crystal switch | |
| CN113392965B (en) | Hadamard product realization method, device and storage medium | |
| CN113673677B (en) | Method, equipment and medium for realizing nonlinear activation function RELU | |
| Wang et al. | Optimal design of a flat-top interleaver based on cascaded M–Z interferometers by using a genetic algorithm | |
| CN113839715A (en) | Optical neural network and chip | |
| Shastri et al. | Integrated neuromorphic photonics | |
| Wei et al. | Inverse design of compact nonvolatile reconfigurable silicon photonic devices with phase-change materials |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |