CN118137277A - Rapid automatic mode locking method, system and equipment based on deep learning - Google Patents
Rapid automatic mode locking method, system and equipment based on deep learning Download PDFInfo
- Publication number
- CN118137277A CN118137277A CN202410545359.0A CN202410545359A CN118137277A CN 118137277 A CN118137277 A CN 118137277A CN 202410545359 A CN202410545359 A CN 202410545359A CN 118137277 A CN118137277 A CN 118137277A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- time domain
- polarization
- mode
- domain waveform
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013135 deep learning Methods 0.000 title claims abstract description 27
- 230000010287 polarization Effects 0.000 claims abstract description 63
- 230000015654 memory Effects 0.000 claims abstract description 17
- 230000007246 mechanism Effects 0.000 claims abstract description 15
- 238000004891 communication Methods 0.000 claims abstract description 5
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 238000013136 deep learning model Methods 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 11
- 239000013598 vector Substances 0.000 claims description 11
- 239000000835 fiber Substances 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 abstract description 10
- 238000006243 chemical reaction Methods 0.000 abstract 1
- 230000003287 optical effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 239000000463 material Substances 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 230000003595 spectral effect Effects 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 229920006395 saturated elastomer Polymers 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 241000282414 Homo sapiens Species 0.000 description 1
- 239000006096 absorbing agent Substances 0.000 description 1
- 239000011358 absorbing material Substances 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012984 biological imaging Methods 0.000 description 1
- 239000002041 carbon nanotube Substances 0.000 description 1
- 229910021393 carbon nanotube Inorganic materials 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000000386 microscopy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002086 nanomaterial Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000005693 optoelectronics Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Landscapes
- Lasers (AREA)
Abstract
The invention discloses a rapid automatic mode locking method, a system and equipment based on deep learning, belonging to the fields of mode locking lasers and automatic control, comprising the following steps: collecting output signals of mode-locked lasers under different angles through an electric steering engine rotating polarization controller; then forming a data set after photoelectric conversion and high-speed sampling for training a model; the training model method adopts a long-period memory model and combines an attention mechanism; after training, the model is used for inputting needed mode-locking time domain waveform data, the adjusted polarization state value is input into a steering engine control board by a computer through a serial port communication protocol through reverse modeling and converted into direct current voltage, and the output direct current voltage drives an electric control steering engine in a mode-locking laser system, so that automatic rotation polarization control is realized. The method solves the problems of long time consumption and insufficient stability of manual adjustment of polarization control in the passive mode-locked laser based on nonlinear polarization evolution.
Description
Technical Field
The invention relates to the field of mode locking lasers and automatic control, in particular to a rapid automatic mode locking method, system and equipment based on deep learning.
Background
Mode-locked lasers have attracted extensive attention in the field of optoelectronic technology and become a key component of ultra-fast optical system research. The technology plays an important role in the subjects of physics, chemistry, biology, material science, information science and the like, and the motion rule of substances in the microscopic world is revealed by generating ultrashort pulses. The mode-locked laser has been widely used in the fields of material processing, microscopy, biological imaging, distance measurement, dimension measurement, clock and synchronization, optical communication, optical signal processing, remote sensing, etc. by its characteristics of high pulse energy and narrow pulse width.
Modes for realizing the mode-locked laser include active mode locking, passive mode locking and hybrid mode locking. In recent years, passive mode-locked fiber lasers have received increasing attention in the fields of telecommunications, metrology, material processing, and the like. Various saturated absorbing materials, such as semiconductor saturable absorbing mirrors, carbon nanotubes, two-dimensional nanomaterials, etc., have been widely studied for activating mode locking. However, these materials often suffer from lower damage thresholds and reduced performance over time.
To address these problems, a method of inducing an artificial saturated absorber mechanism based on nonlinear optical effects is attracting attention. Including nonlinear polarization rotation (NPE), these artificial sensors have the advantage of being independent of the laser operating wavelength and easy to implement. However, in setting nonlinear polarization rotation, the polarization state needs to be precisely adjusted, which is easily affected by environmental interference and has poor stability. To overcome these problems, some experiments have attempted to achieve fast auto-mode locking using electronically controlled polarization techniques. However, the existing method still has the problems of complex structure, difficult identification, long time consumption and the like.
Disclosure of Invention
In order to solve the defects in the background art, the invention aims to provide a rapid automatic mode locking method, a system and equipment based on deep learning, which can realize automatic mode locking based on the polarization control problem of a nonlinear polarization rotation mode locking fiber laser.
In a first aspect, the object of the present invention can be achieved by the following technical solutions: a fast automatic mode locking method based on deep learning comprises the following steps:
receiving a time domain waveform signal, and generating a data set according to the time domain waveform signal, wherein the time domain waveform signal is obtained by photoelectrically converting an output signal of a nonlinear polarization rotation mode-locked fiber laser under different polarization controller angles;
inputting the data set into a pre-established deep learning model, outputting to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the acquisition process of the time domain waveform signal utilizes a photoelectric detector to acquire data, and then the data is processed by a high-speed oscilloscope to obtain the time domain waveform signal.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the high-speed sampling of the time domain waveform signals realizes the digitization of the output signals of the mode-locked laser through the function of storing files by a spectrometer.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the mode locking state of the mode locking time domain waveform data comprises Shan Guzi mode locking and automatic identification of a Q-switched mode locking state.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the pre-established deep learning model expression is combined with a long-period and short-period memory model of an attention mechanism, and the principle of the attention mechanism is as follows:
Input data Where n is the sequence length,/>Representing the i-th element of the input sequence. First, the attention score is calculated:
Wherein d is a normalization factor, ，/>，/>And/>Is a learnable weight matrix,/>Query vectors representing the ith element in the input sequence for measuring importance of that element to other elements,/>A key vector representing the j-th element in the input sequence, for measuring the importance of other elements to the element;
And secondly, calculating the attention weight:
Wherein, Converting the value of each element in the score vector to a value between 0 and 1 for a normalization function,/>Representing the attention weight between the i-th element and the j-th element in the input sequence.
Attention weights are then added to the sequence values:
Wherein, ，/>Is a learnable weight matrix,/>Is a weighted representation of the i-th element in the input sequence by the attention mechanism. Wherein/>、/>、/>Is the model parameter that needs to be learned.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the process of realizing the rotation polarization control according to the angles of the two polarization controllers comprises the following steps:
the obtained angle values of the two polarization controllers are input into the singlechip through an SSH communication protocol by a computer, are converted into direct-current voltage by the singlechip, and the direct-current voltage output by the singlechip drives the steering engine control board, so that automatic rotary polarization control is realized.
With reference to the first aspect, in certain implementations of the first aspect, the method further includes: the steering engine control board drives the steering engine to rotate the polarization controller by using the control voltage of 0-6V, so that all polarization states are generated.
In a second aspect, to achieve the above object, the present invention discloses a fast automatic mode locking system based on deep learning, including:
The data generation module is used for receiving the time domain waveform signals and generating a data set according to the time domain waveform signals, wherein the time domain waveform signals are obtained by photoelectrically converting output signals of the nonlinear polarization rotation mode-locked fiber laser under different angles of the polarization controller;
The polarization control module is used for inputting the data set into a pre-established deep learning model, outputting the data set to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting the data set to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
In another aspect of the present invention, in order to achieve the above object, a terminal device is disclosed, which includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores the computer program capable of running on the processor, and when the processor loads and executes the computer program, a fast automatic mode locking method based on deep learning as described above is adopted.
In a further aspect of the present invention, in order to achieve the above object, a computer readable storage medium is disclosed, in which a computer program is stored, which when loaded and executed by a processor, employs a fast automatic mode locking method based on deep learning as above.
The invention has the beneficial effects that:
According to the invention, by utilizing a long-period memory model method based on a combined attention mechanism and combining mode locking state pulse identification prediction, automatic polarization control and quick automatic mode locking are realized, wherein the mode locking of an automatic single soliton is realized at the fastest time only for 6.1s, and the polarization state can be changed by inputting different mode locking state diagrams, so that the multi-state quick switching is realized. The method solves the problem of polarization control in the passive mode-locking laser, and can enable the mode-locking laser to automatically and rapidly lock the mode.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described, and it will be obvious to those skilled in the art that other drawings can be obtained according to these drawings without inventive effort;
FIG. 1 is a schematic diagram of a short-term memory model based on a combined attention mechanism and including rapid recognition and search of mode-locked pulses in an embodiment of the present invention;
FIG. 2 is a flow chart of a method in an embodiment of the invention;
FIG. 3 is a graph showing the relationship between training loss value and training frequency in the embodiment of the present invention;
FIG. 4 is a schematic diagram of an implementation of an auto-mode-locked fiber laser;
FIG. 5 is a schematic diagram of a deep learning model according to an embodiment of the present invention;
FIG. 6 is a comparison of a test spectral plot with a prediction in an embodiment of the present invention;
FIG. 7 is a diagram of a fast identification Shan Guzi mode-locking spectrum in an actual optical path in an embodiment of the present invention;
FIG. 8 is a schematic diagram of the system of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Embodiment one:
the following description is made of the relevant terms related to the embodiments of the present application:
deep learning: deep learning (DL, deep Learning) is a new research direction in the field of machine learning (ML, machine Learning) that was introduced into machine learning to bring it closer to the original goal-artificial intelligence (AI, artificial Intelligence).
Deep learning is the inherent regularity and presentation hierarchy of learning sample data, and the information obtained during such learning is helpful in interpreting data such as text, images and sounds. Its final goal is to have the machine have analytical learning capabilities like a person, and to recognize text, image, and sound data. Deep learning is a complex machine learning algorithm that achieves far greater results in terms of speech and image recognition than prior art.
Deep learning has achieved many results in search technology, data mining, machine learning, machine translation, natural language processing, multimedia learning, speech, recommendation, and personalization technologies, as well as other related fields. The deep learning makes the machine imitate the activities of human beings such as audio-visual and thinking, solves a plurality of complex pattern recognition problems, and makes the related technology of artificial intelligence greatly advanced.
As shown in fig. 1, a fast automatic mode locking method based on deep learning includes the following steps:
receiving a time domain waveform signal, and generating a data set according to the time domain waveform signal, wherein the time domain waveform signal is obtained by photoelectrically converting an output signal of a nonlinear polarization rotation mode-locked fiber laser under different polarization controller angles;
the time domain waveform signal is obtained by collecting data by using a photoelectric detector and then processing the data by using a high-speed oscilloscope;
further, the high-speed sampling of the time domain waveform signal is realized by the file storage function of the spectrometer
Further, the data set is divided into a test set and a training set, a deep learning model is trained, and the relation between the training loss value and the training times is shown in a figure 3;
inputting the data set into a pre-established deep learning model, outputting to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
Further, the mode locking state of the mode locking time domain waveform data comprises Shan Guzi automatic identification of the mode locking state and the Q-switched mode locking state;
in lasers, the laser output often has multiple modes due to mode competition and nonlinear effects within the optical cavity, and the fundamental mode is the lowest frequency mode in the laser output.
Fundamental mode locking means that the output of the laser remains stable in fundamental mode without frequency drift or frequency hopping. This is important for many applications, especially for experiments and applications where precise control of the laser frequency is required. The frequency stability of the output of the laser can be ensured through fundamental frequency mode locking, so that the accuracy and the repeatability of the experiment are improved.
Wherein the pre-established deep learning model is a long-term and short-term memory model combined with an attention mechanism;
The bonding process is as follows: firstly, in the custom attention layer, the input feature vector is mapped through 3 linear layers and mapped into the query, key and value spaces respectively, and in the forward propagation function, the dot product between the query and the key is calculated by using a softmax function to calculate the attention weight. This result is then divided by the square root of the query dimension to determine the attention score. These scores were then normalized using the softmax function to obtain the attention weight. The attention layer is integrated into the LSTM model. The output of the LSTM layer is taken as input and the attention mechanism is utilized to dynamically focus on the various segments of the input sequence. This enables the model to more effectively capture critical information in the sequence.
In a specific LSTM model setting, the input dimension of the PC corner is 2 and the output dimension is 1800. The overall architecture consists of 3 LSTM layers and 2 fully connected layers. Wherein, the number of hidden units of the three LSTM layers is 32, 256 and 512, and the number of hidden units of the two full connection layers is 512 and 1024. Increasing the number of hidden units layer by layer helps the model learn and extract more complex features and temporal relationships in the input data step by step. The activation function utilizes ReLU6 to enhance the nonlinear fitting ability of the model and applies the softmax function to the final fully connected layer for output. Matching the hidden state dimension of the attention mechanism with the output dimension of the last LSTM layer ensures that the attention layer can effectively capture important information in the hidden state. In addition, the Adam optimizer is selected for parameter update and the learning rate is set to 0.0005 to adjust the step size of the parameter update.
The pre-established long-period memory model combined with the attention mechanism has the following principle:
Input data Where n is the sequence length,/>Representing the i-th element of the input sequence. First, the attention score is calculated:
Wherein d is a normalization factor, ，/>，/>And/>Is a learnable weight matrix. /(I)Query vectors representing the ith element in the input sequence for measuring importance of that element to other elements,/>The key vector representing the j-th element in the input sequence is used to measure the importance of other elements to that element.
And secondly, calculating the attention weight:
Wherein, Converting the value of each element in the score vector to a value between 0 and 1 for a normalization function,/>Representing the attention weight between the i-th element and the j-th element in the input sequence.
Attention weights are then added to the sequence values:
Wherein, ，/>Is a learnable weight matrix,/>Is a weighted representation of the i-th element in the input sequence by the attention mechanism. Wherein/>、/>、/>Is the model parameter that needs to be learned.
Specifically, the following examples are provided to further illustrate the present invention:
After training, predicting a mode-locking spectrum image, if the prediction result has good effect, inputting mode-locking time domain waveform data into a trained deep learning model, and enabling the model to output angles of two polarization controllers corresponding to the fundamental frequency mode-locking image by using a back propagation algorithm of the trained deep learning model, wherein a specific flow chart of the process is shown in fig. 4.
Further, the angle value obtained through training is input into the singlechip through an SSH communication protocol by a computer, the angle value is converted into direct-current voltage through the singlechip, and the direct-current voltage output by the singlechip drives the steering engine control board, so that automatic rotary polarization control is realized.
The steering engine control board can drive the steering engine to rotate a Polarization Controller (PC) by using a control voltage of 0-6V, so that all polarization states can be generated;
FIG. 2 is a flow chart of a method in an embodiment of the invention. It should be noted that noise reduction is required after the spectral data is acquired, and a frequency domain filtering mode is used herein. The method comprises the following specific steps:
Step 1: fourier transform is one method of converting a signal from the time domain to the frequency domain. The method comprises the following specific steps:
first, the raw spectral data is represented as a one-dimensional array or vector.
The array is then converted from the time domain to the frequency domain using a fast fourier transform.
Step 2: a frequency axis is acquired. The frequency axis represents the distribution of spectral data over the frequency domain.
Step 3: threshold filtering (the filtering method used is the Savitzky-Golay algorithm) is used to smooth and reduce noise in the time series data. The SG algorithm works by sliding a window over the data, fitting a polynomial to the points in the window at each position, and replacing it with the value estimated by the fitting polynomial, effectively smoothing the noise in the spectrum.
Step 4: and performing inverse Fourier transform on the data after noise reduction. It is converted back into the time domain.
Embodiment two: in a second aspect, as shown in fig. 8, an embodiment of the present application discloses a fast automatic mode locking system based on deep learning, including
The data generation module is used for receiving the time domain waveform signals and generating a data set according to the time domain waveform signals, wherein the time domain waveform signals are obtained by photoelectrically converting output signals of the nonlinear polarization rotation mode-locked fiber laser under different angles of the polarization controller;
The polarization control module is used for inputting the data set into a pre-established deep learning model, outputting the data set to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting the data set to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
Based on the same inventive concept, the present invention also provides a computer apparatus comprising: one or more processors, and memory for storing one or more computer programs; the program includes program instructions and the processor is configured to execute the program instructions stored in the memory. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processor, digital signal processor (DIGITAL SIGNAL Processor, DSP), application specific integrated circuit (Application SpecificIntegrated Circuit, ASIC), field-Programmable gate array (Field-Programmable GATEARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc., that are the computational core and control core of the terminal for implementing one or more instructions, particularly for loading and executing one or more instructions within a computer storage medium to implement the methods described above.
It should be further noted that, based on the same inventive concept, the present invention also provides a computer storage medium having a computer program stored thereon, which when executed by a processor performs the above method. The storage media may take the form of any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
In the description of the present specification, the descriptions of the terms "one embodiment," "example," "specific example," and the like, mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The foregoing has shown and described the basic principles, principal features, and advantages of the present disclosure. It will be understood by those skilled in the art that the present disclosure is not limited to the embodiments described above, which have been described in the foregoing and description merely illustrates the principles of the disclosure, and that various changes and modifications may be made therein without departing from the spirit and scope of the disclosure, which is defined in the appended claims.
Claims (10)
1. The quick automatic mode locking method based on deep learning is characterized by comprising the following steps of:
receiving a time domain waveform signal, and generating a data set according to the time domain waveform signal, wherein the time domain waveform signal is obtained by photoelectrically converting an output signal of a nonlinear polarization rotation mode-locked fiber laser under different polarization controller angles;
inputting the data set into a pre-established deep learning model, outputting to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
2. The method for rapid automatic mode locking based on deep learning according to claim 1, wherein the time domain waveform signal acquisition process uses a photodetector to acquire data, and the data is processed by a high-speed oscilloscope to obtain the time domain waveform signal.
3. The fast automatic mode locking method based on deep learning according to claim 2, wherein the high-speed sampling of the time domain waveform signal realizes the digitization of the output signal of the mode locking laser through a file storage function of a spectrometer.
4. The method of claim 1, wherein the mode locking state of the mode locking time domain waveform data comprises automatic identification of Shan Guzi mode locking and Q-switched mode locking states.
5. The fast automatic mode locking method based on deep learning according to claim 1, wherein the pre-established deep learning model uses a long-short term memory model combined with a attention mechanism, and the principle of the attention mechanism is as follows:
Input data Where n is the sequence length,/>An i-th element representing the input sequence, first calculates the attention score:
Wherein d is a normalization factor, ，/>，/>And/>Is a learnable weight matrix,/>Query vector representing the i-th element in the input sequence,/>A key vector representing a j-th element in the input sequence;
And secondly, calculating the attention weight:
Wherein, Converting the value of each element in the score vector to a value between 0 and 1 for a normalization function,/>Representing an attention weight between an ith element and a jth element in the input sequence;
attention weights are then added to the sequence values:
Wherein, ，/>Is a learnable weight matrix,/>Is a representation of the i-th element in the input sequence weighted by the attention mechanism, where/>、/>、/>Is the model parameter that needs to be learned.
6. The fast auto-mode locking method based on deep learning according to claim 1, wherein the process of implementing the rotation polarization control according to the angles of two polarization controllers:
the obtained angle values of the two polarization controllers are input into the singlechip through an SSH communication protocol by a computer, are converted into direct-current voltage by the singlechip, and the direct-current voltage output by the singlechip drives the steering engine control board, so that automatic rotary polarization control is realized.
7. The fast auto-mode locking method based on deep learning according to claim 6, wherein the steering engine control board generates all polarization states by generating a steering engine rotation polarization controller using a control voltage of 0-6V.
8. A quick automatic mode locking system based on deep learning is characterized by comprising
The data generation module is used for receiving the time domain waveform signals and generating a data set according to the time domain waveform signals, wherein the time domain waveform signals are obtained by photoelectrically converting output signals of the nonlinear polarization rotation mode-locked fiber laser under different angles of the polarization controller;
The polarization control module is used for inputting the data set into a pre-established deep learning model, outputting the data set to obtain a trained deep learning model, receiving mode-locked time domain waveform data, inputting the mode-locked time domain waveform data into the trained deep learning model, outputting the data set to obtain angles of two polarization controllers, and realizing rotary polarization control according to the angles of the two polarization controllers.
9. A terminal device comprising a memory, a processor and a computer program stored in the memory and capable of running on the processor, characterized in that the memory stores the computer program capable of running on the processor, and that the processor adopts a fast automatic mode locking method based on deep learning according to any one of claims 1 to 7 when loading and executing the computer program.
10. A computer readable storage medium having a computer program stored therein, wherein the computer program, when loaded and executed by a processor, employs a fast auto-mode locking method based on deep learning according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410545359.0A CN118137277A (en) | 2024-05-06 | 2024-05-06 | Rapid automatic mode locking method, system and equipment based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410545359.0A CN118137277A (en) | 2024-05-06 | 2024-05-06 | Rapid automatic mode locking method, system and equipment based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118137277A true CN118137277A (en) | 2024-06-04 |
Family
ID=91243181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410545359.0A Pending CN118137277A (en) | 2024-05-06 | 2024-05-06 | Rapid automatic mode locking method, system and equipment based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118137277A (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108539571A (en) * | 2018-04-08 | 2018-09-14 | 上海交通大学 | A kind of fast automatic mode locking method covering multimode pulse recognition |
| CN108764303A (en) * | 2018-05-10 | 2018-11-06 | 电子科技大学 | A kind of remote sensing images spatial term method based on attention mechanism |
| CN110299666A (en) * | 2019-06-11 | 2019-10-01 | 上海交通大学 | A kind of anthropomorphic algorithm for mode locking automatic under a variety of pulse conditions |
| CN112234423A (en) * | 2020-10-14 | 2021-01-15 | 江苏师范大学 | Dual-wavelength mode-locked fiber laser with wavelength interval exceeding 50 nanometers and dual-wavelength mode-locked laser output generation method |
| CN113300198A (en) * | 2021-05-14 | 2021-08-24 | 江苏师范大学 | Single-frequency narrow-linewidth distributed feedback pulse fiber laser and implementation method thereof |
| CN114034653A (en) * | 2021-11-17 | 2022-02-11 | 清华大学 | Wavelength modulation absorption spectrum chromatography reconstruction system based on deep learning |
| CN114818488A (en) * | 2022-04-20 | 2022-07-29 | 上海交通大学 | Ultrafast pulse laser reverse implementation method and system based on reinforcement learning |
| CN116435861A (en) * | 2023-04-14 | 2023-07-14 | 中红外激光研究院(江苏)有限公司 | Device and method for automatically identifying and switching state of middle-infrared ultrashort pulse laser |
| CN116798004A (en) * | 2022-09-09 | 2023-09-22 | 南京信息工程大学 | Vehicle re-identification method, system, device and storage medium |
| CN117081831A (en) * | 2023-09-07 | 2023-11-17 | 南京信息工程大学 | Network intrusion detection method and system based on data generation and attention mechanism |
| CN117236414A (en) * | 2023-08-30 | 2023-12-15 | 南京理工大学 | Laser cavity stable mode locking method of mode locking laser based on reinforcement learning |
| CN117595058A (en) * | 2023-10-07 | 2024-02-23 | 四川大学 | Intelligent light-emitting and mode locking system based on SESAM mode locking solid laser |
| CN117791286A (en) * | 2023-12-27 | 2024-03-29 | 中红外激光研究院(江苏)有限公司 | Mid-infrared ultrafast laser control system and method based on data-driven neural network |
-
2024
- 2024-05-06 CN CN202410545359.0A patent/CN118137277A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108539571A (en) * | 2018-04-08 | 2018-09-14 | 上海交通大学 | A kind of fast automatic mode locking method covering multimode pulse recognition |
| CN108764303A (en) * | 2018-05-10 | 2018-11-06 | 电子科技大学 | A kind of remote sensing images spatial term method based on attention mechanism |
| CN110299666A (en) * | 2019-06-11 | 2019-10-01 | 上海交通大学 | A kind of anthropomorphic algorithm for mode locking automatic under a variety of pulse conditions |
| CN112234423A (en) * | 2020-10-14 | 2021-01-15 | 江苏师范大学 | Dual-wavelength mode-locked fiber laser with wavelength interval exceeding 50 nanometers and dual-wavelength mode-locked laser output generation method |
| CN113300198A (en) * | 2021-05-14 | 2021-08-24 | 江苏师范大学 | Single-frequency narrow-linewidth distributed feedback pulse fiber laser and implementation method thereof |
| CN114034653A (en) * | 2021-11-17 | 2022-02-11 | 清华大学 | Wavelength modulation absorption spectrum chromatography reconstruction system based on deep learning |
| CN114818488A (en) * | 2022-04-20 | 2022-07-29 | 上海交通大学 | Ultrafast pulse laser reverse implementation method and system based on reinforcement learning |
| CN116798004A (en) * | 2022-09-09 | 2023-09-22 | 南京信息工程大学 | Vehicle re-identification method, system, device and storage medium |
| CN116435861A (en) * | 2023-04-14 | 2023-07-14 | 中红外激光研究院(江苏)有限公司 | Device and method for automatically identifying and switching state of middle-infrared ultrashort pulse laser |
| CN117236414A (en) * | 2023-08-30 | 2023-12-15 | 南京理工大学 | Laser cavity stable mode locking method of mode locking laser based on reinforcement learning |
| CN117081831A (en) * | 2023-09-07 | 2023-11-17 | 南京信息工程大学 | Network intrusion detection method and system based on data generation and attention mechanism |
| CN117595058A (en) * | 2023-10-07 | 2024-02-23 | 四川大学 | Intelligent light-emitting and mode locking system based on SESAM mode locking solid laser |
| CN117791286A (en) * | 2023-12-27 | 2024-03-29 | 中红外激光研究院(江苏)有限公司 | Mid-infrared ultrafast laser control system and method based on data-driven neural network |
Non-Patent Citations (2)
| Title |
|---|
| FANG, Y 等: "Deep neural network for modeling soliton dynamics in the mode-locked laser", 《OPTICS LETTERS》, vol. 48, no. 3, 10 August 2023 (2023-08-10), pages 779 - 782 * |
| 罗仪豪 等: "基于深度学习的超材料涉及及光纤光束控制研究进展", 《中国激光》, vol. 50, no. 11, 15 June 2023 (2023-06-15), pages 104 - 122 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pereira et al. | Unsupervised anomaly detection in energy time series data using variational recurrent autoencoders with attention | |
| Rafiq et al. | Regularized LSTM based deep learning model: first step towards real-time non-intrusive load monitoring | |
| CN107909227B (en) | Method and device for ultra-short-term prediction of wind power plant power and wind generating set | |
| JP2023523726A (en) | Supervised Contrastive Learning with Multiple Positive Examples | |
| CN112435142B (en) | Power load identification method and load power utilization facility knowledge base construction method thereof | |
| CN111598303B (en) | Summer short-term load prediction method based on meteorological component decomposition | |
| An et al. | Speech Emotion Recognition algorithm based on deep learning algorithm fusion of temporal and spatial features | |
| Zhao et al. | A short-term hybrid wind power prediction model based on singular spectrum analysis and temporal convolutional networks | |
| Zheng et al. | Recognition method of voltage sag causes based on two‐dimensional transform and deep learning hybrid model | |
| Cai et al. | Wind power forecasting based on ensemble empirical mode decomposition with generalized regression neural network based on cross-validated method | |
| Guan et al. | Medical image fusion algorithm based on multi-resolution analysis coupling approximate spare representation | |
| Du et al. | Multi-Domain Kernel Dictionary Learning Sparse Classification Method for Intelligent Machinery Fault Diagnosis | |
| CN118137277A (en) | Rapid automatic mode locking method, system and equipment based on deep learning | |
| CN116543289B (en) | Image description method based on encoder-decoder and Bi-LSTM attention model | |
| Kemeth et al. | Initializing LSTM internal states via manifold learning | |
| CN113074813B (en) | Terahertz time-domain spectroscopy system and step length self-adaptive adjustment method thereof | |
| Jiang et al. | Remote Sensing Images Data Augmentation Based on Style Transfer under the Condition of Few Samples | |
| Yu et al. | Power quality data processing method based on a distributed compressed sensing and learning dictionary | |
| Shi et al. | Recurrent neural network wind power prediction based on variational modal decomposition improvement | |
| Zhang et al. | A CPM signal denoising method based on attention network | |
| Luo | Research on key technologies of speech recognition based on deep learning | |
| Zhao et al. | Short term wind power interval prediction based on vmd and improved bls | |
| Song | Performance of Autoencoder with Bi-Directional Long-Short Term Memory Network in Gestures Unit Segmentation | |
| Wu et al. | Aircraft tracking in infrared imagery with adaptive learning and interference suppression | |
| Yin et al. | Graph attention-based U-net conditional generative adversarial networks for the identification of synchronous generation unit parameters |
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 |