CN117192865A - Spectrum programmable optical frequency comb generation method based on deep reinforcement learning - Google Patents
Spectrum programmable optical frequency comb generation method based on deep reinforcement learning Download PDFInfo
- Publication number
- CN117192865A CN117192865A CN202311082437.XA CN202311082437A CN117192865A CN 117192865 A CN117192865 A CN 117192865A CN 202311082437 A CN202311082437 A CN 202311082437A CN 117192865 A CN117192865 A CN 117192865A
- Authority
- CN
- China
- Prior art keywords
- spectrum
- frequency comb
- optical frequency
- reinforcement learning
- deep reinforcement
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000003287 optical effect Effects 0.000 title claims abstract description 86
- 238000001228 spectrum Methods 0.000 title claims abstract description 80
- 230000002787 reinforcement Effects 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims description 27
- 230000006870 function Effects 0.000 claims abstract description 35
- 230000009471 action Effects 0.000 claims abstract description 28
- 230000003993 interaction Effects 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 8
- 239000003795 chemical substances by application Substances 0.000 claims description 31
- 239000013307 optical fiber Substances 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 8
- 238000002474 experimental method Methods 0.000 claims description 7
- 230000001276 controlling effect Effects 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 6
- 238000007493 shaping process Methods 0.000 claims description 6
- 230000009022 nonlinear effect Effects 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 4
- 239000000835 fiber Substances 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 4
- 230000008901 benefit Effects 0.000 claims description 2
- 230000010287 polarization Effects 0.000 claims description 2
- 230000001105 regulatory effect Effects 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 7
- 238000004891 communication Methods 0.000 abstract description 3
- 238000005259 measurement Methods 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 238000005457 optimization Methods 0.000 description 5
- 230000002452 interceptive effect Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 210000001520 comb Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Landscapes
- Optical Communication System (AREA)
Abstract
A spectrum programmable optical frequency comb generating system based on deep reinforcement learning deploys an intelligent body of deep reinforcement learning based on an experimental physical structure of a broadband optical frequency comb and constructs an interaction model between the intelligent body and an experimental environment; constructing a strategy algorithm framework based on a deep reinforcement learning Actor-Critic architecture, and designing interaction content and rules between an intelligent agent and an experimental environment according to the strategy algorithm framework; and constructing a reward function by taking root mean square error of a target frequency spectrum and an experimental frequency spectrum as parameters, designing action execution and reward feedback of an agent and an experimental environment, and obtaining an optimal phase modulation decision by a training strategy of a deep reinforcement learning algorithm so as to realize the programmable generation of an optical frequency comb by the spectrum. The invention trains the neural network by using the deep reinforcement learning technology, selects the optimal phase modulation strategy, and realizes the spectrum programming and control of the optical frequency comb. The application range of the optical frequency comb is widened, and higher flexibility is provided for the application of the optical frequency comb in the fields of optical communication, precise measurement and the like.
Description
Technical Field
The invention relates to a technology in the field of optical signal processing, in particular to a spectrum programmable optical frequency comb generating method based on deep reinforcement learning.
Background
The existing optical frequency comb generation technology comprises novel technologies such as an optical frequency comb based on nonlinear optical effect and an optical frequency comb based on microwave optical mixing, but the technologies have problems of frequency loss, phase jitter, limited tuning range and the like, and the application of the technologies in the fields of high-precision measurement, optical communication and the like is affected.
Through the search of the prior art, tingyang et al published in 2013 Optics express, volume 21, 7, paper "Comparison analysis ofoptical frequency comb generationwith nonlinear effects in highly nonlinear fibers" proposed a broadband optical frequency comb generation scheme based on cascaded four-wave mixing and self-phase modulation. The proposal realizes 259-line optical frequency comb with the repetition rate of 10GHz and the flatness within 5dB by using two cascaded high-nonlinearity optical fibers with different zero dispersion wavelengths. Although this scheme enables independent tuning of the repetition rate and center frequency of the optical frequency comb, it does not support control and shaping of the frequency spectrum of the optical frequency comb.
Whereas the paper "Enhanced nonlinear spectral broadening and sub-picosecond pulse generation by adaptive spectral phase optimization of electro-optic frequency combs" published by Vikram et al in 2020 Optics Express, volume 28, phase 8, proposes a broadband optical frequency comb generation method based on an adaptive optimization algorithm. According to the scheme, the frequency spectrum phase of the electro-optic frequency comb is adaptively adjusted through the Fourier pulse shaper to improve the stimulated Brillouin scattering threshold, and the bandwidth of the optical frequency comb can be increased to be more than 13 times. However, the widening of the optical frequency comb is a complex nonlinear process, and by the scheme, only a relatively fuzzy optimization target can be processed, and the frequency spectrum of the optical frequency comb cannot be accurately controlled.
Disclosure of Invention
Aiming at the defect that the prior art is difficult to manufacture a target optical frequency comb completely according to simulation data and the optical frequency comb manufactured based on a micro-ring resonant cavity platform cannot dynamically regulate and control the frequency spectrum in real time according to requirements, the invention provides a spectrum programmable optical frequency comb generating system based on deep reinforcement learning, which utilizes the deep reinforcement learning technology to train a neural network and selects an optimal phase modulation strategy to realize spectrum programming and control of the optical frequency comb. The application range of the optical frequency comb is widened, and higher flexibility is provided for the application of the optical frequency comb in the fields of optical communication, precise measurement and the like.
The invention is realized by the following technical scheme:
the invention relates to a spectrum programmable optical frequency comb generating method based on deep reinforcement learning, which deploys an intelligent body of the deep reinforcement learning based on an experimental physical structure of a broadband optical frequency comb and constructs an interaction model between the intelligent body and an experimental environment; constructing a strategy algorithm framework based on a deep reinforcement learning Actor-Critic architecture, and designing interaction content and rules between an intelligent agent and an experimental environment according to the strategy algorithm framework; and constructing a reward function by taking root mean square error of a target frequency spectrum and an experimental frequency spectrum as parameters, designing action execution and reward feedback of an agent and an experimental environment, and obtaining an optimal phase modulation decision by a training strategy of a deep reinforcement learning algorithm so as to realize the programmable generation of an optical frequency comb by the spectrum.
Technical effects
According to the invention, an interactive model of an intelligent body and an optical frequency comb generating system environment is constructed by utilizing deep reinforcement learning, an initial optical frequency comb is widened through a nonlinear effect of a high nonlinear optical fiber, and a training strategy of a deep reinforcement learning algorithm is designed to construct a reward function, so that an optimal phase modulation decision is output. The invention applies the deep reinforcement learning to the optical frequency comb control and shaping of the closed-loop optical system, thereby realizing the broadband optical frequency comb with programmable frequency spectrum. Compared with the prior art, the technical effect of the invention is obviously superior in a plurality of aspects: firstly, the spectrum programmability enables the system to generate customized spectrum according to specific requirements, and further expands the application field of the broadband optical frequency comb. And secondly, through training of the intelligent body based on the experimental environment, the optimization decision of the system has robustness and is not influenced by environmental noise, so that the reliability of the system in a complex environment is ensured. In addition, through the optimization decision of the intelligent agent, the system can realize the spectrum shaping and control of the optical frequency comb on the premise of unchanged experimental structure, and the working hour cost is obviously reduced.
Drawings
FIG. 1 is a block diagram of a system of the present invention;
FIG. 2 is a flow chart of an embodiment;
deep reinforcement learning network structure diagram of the embodiment of fig. 3.
Detailed Description
As shown in fig. 1, this embodiment relates to a machine learning-based spectrum-designable optical frequency comb generating system, which includes: the laser, polarization controller, intensity modulator, phase modulator, programmable light processor, ooze bait fiber amplifier, nonlinear optical fiber and spectrum appearance that link to each other in proper order, wherein: the intensity modulator and the phase modulator control modulation parameters through the radio frequency source and the phase shifter, the spectrometer generates a spectrum initial state and a spectrum execution state according to the acquired broadened spectrum information, and the intelligent agent is used for controlling the phase of the programmable light processor.
As shown in fig. 2, this embodiment relates to an optical frequency comb generating method based on the above system, including:
step A, constructing a stable electro-optic optical frequency comb, regulating and controlling spectrum phase, and widening the spectrum of the optical frequency comb by utilizing nonlinear effect, wherein the method specifically comprises the following steps of:
a1, generating a stable initial electro-optic frequency comb by a single wavelength laser, one intensity modulator and two phase modulators in cascade, wherein: the single wavelength laser outputs input light with 1550nm wavelength, the power is 10dBm, the center frequency of the generated initial optical frequency comb is 193.548THz, and the repetition rate is 10GHz.
Preferably, the direct-current bias voltage is loaded on the intensity modulator to enable the intensity modulator to work at the right-angle point, so that an initial optical frequency comb with flat frequency spectrum and more comb teeth is generated.
A2, modulating the phase of the initial optical frequency comb by using a programmable optical processor: selecting a phase within 8nm bandwidth near the center frequency of the initial optical frequency comb, namely 1546nm to 1554nm, as a modulation object, and using the phase modulation curve as a weightThe weighted sum of the heavy random 20 th order chebyshev polynomials is expressed as follows: chebyshev polynomial function T with nth power n (x)＝cos(ncos -1 (x))，Wherein: w (w) k Representing chebyshev polynomials T k (x) Weight, W of (2) 20 (x) Representing the result of the weighted addition of chebyshev polynomials.
A3, inputting the modulated electro-optical frequency comb into an erbium-doped optical fiber amplifier, injecting the electro-optical frequency comb into a high-nonlinearity optical fiber, and utilizing a nonlinearity effect to carry out optical frequency comb spectrum broadening.
Preferably, the modulated initial optical frequency comb is amplified to 23dBm and then injected into a high nonlinearity optical fiber with 2km length for nonlinear broadening, and the nonlinearity parameter is 0.03 ps/(nm) 2 ·km)。
And A4, acquiring and expanding spectrum information by using a spectrometer, and analyzing to obtain a spectrum initial state and a spectrum execution state.
And B, based on the physical structure of the broadband optical frequency comb obtained through experiments, namely the spectrum initial state and the spectrum execution state obtained in the step A, deploying the intelligent body for deep reinforcement learning, and establishing an interaction model of the intelligent body and an experimental environment. As shown in fig. 3, a policy algorithm framework based on a deep reinforcement learning Actor-Critic architecture is established, and interactive contents and rules of a deep reinforcement learning agent and an experimental environment module are designed, which specifically include:
b1, constructing a deep reinforcement learning agent module based on an Actor-Critic architecture, wherein: the Actor network generates an action strategy, such as outputting a phase modulation decision, and updates the action strategy according to a cost function Q feedback provided by the Critic network so as to improve the performance of the intelligent agent in the environment; the Critic network evaluates the merits of the action strategies and calculates a cost function Q according to the actions taken by the agent to evaluate the merits of the current action strategies, and stores the state transition process after the Actor network interacts with the environment in an experience playback memory.
B2, training the intelligent agent through a network random sampling experience playback memory, and specifically comprises the following steps:
(1) setting a reward function R in view of the Root Mean Square Error (RMSE) of the target spectrum and the experimental spectrum being a key indicator for assessing the control and shaping effects of the optical comb t ＝-RMSE(S target ，S exp ) Wherein: target frequency spectrum S target Experiment frequency spectrum S exp The negative sign is to convert RMSE into a maximization problem, and the task of the agent becomes to minimize the difference between the target spectrum and the experimental spectrum;
(2) the goal of setting the agent is to learn an optimal policy function pi (a t |s t θ), wherein: a, a t Action s generated for agent through Actor network t The state of the experimental environment is that theta is a parameter of an Actor network; policy gradients in an Actor networkWherein: a(s) t ,a t ) Is a dominance function of the Critic network; calculating a loss function->Wherein: e represents an expected value, KL represents KL divergence, θ old Representing old Actor network parameters, lambda is a super parameter for controlling the balance of policy update and policy stability;
(3) evaluating the merits of action strategies generated by an Actor network through a Critic network: setting a cost function Q(s) of a Critic network t ,a t ) I.e. in state s t Take action a t Updating using a method based on TD error with the dominant function: delta t ＝r t +γQ(s t+1 ,a t+1 )-Q(s t ,a t ) Dominance function A (s t ,a t )＝Q(s t ,a t )-V(s t ) Wherein: v(s) t ) As a state value function of the Critic network, the variance is reduced and the convergence speed of the algorithm is improved by adopting the advantage function.
B3, adding a small Gaussian noise to the output of the Actor network at each time step to encourage intelligenceThe energy explores new action strategies and avoids over-reliance on past experience. Specifically, for the generated action a t The added gaussian noise obeys a gaussian distribution with zero mean and sigma standard deviation: a' t ＝a t +n (0, σ), wherein: a, a t Representing the original action generated by the agent at time step t, i.e., the phase modulation decision, which is the output of the Actor network; a' t Representing an action after adding gaussian noise, which is to be performed in a real environment; n (0, σ) represents a gaussian distribution with zero mean and σ standard deviation.
The intelligent agent model is trained to obtain an optimal strategy function pi (a) through rewarding functions and interaction with experimental environment t |s t θ). According to the optimal phase control decision of the intelligent body output, the modulated initial optical frequency comb has a target frequency spectrum defined by user programming through a broadened optical frequency comb obtained by a high-nonlinearity optical fiber.
Through specific practical experiments, the spectrum programmable optical frequency comb generating system realizes a target spectrum with user programming definition. The types of spectrum that can be realized can be described by Gaussian functions and Gaussian mixture models with different characteristic parameters:wherein: s (f) represents the generated target spectrum, N represents the number of Gaussian functions, A i 、f i Sum sigma i The amplitude, center frequency and standard deviation of the ith gaussian function are shown, respectively.
According to the invention, by utilizing an interactive model of a deep reinforcement learning intelligent body and an experimental environment and combining the nonlinear effect of the high nonlinear optical fiber with intelligent phase regulation, the precise control of the optical frequency comb broadening spectrum is realized. Compared with the prior art, the method can process complex nonlinear processes, is not influenced by process errors in the manufacturing process of the optical frequency comb, and can ensure accurate generation of the target frequency spectrum. And secondly, the system adopts a strategy algorithm framework based on a deep reinforcement learning Actor-Critic framework, and can dynamically regulate and control the frequency spectrum in real time in the closed loop experiment process through the interactive content and rule design of the intelligent body and the experiment environment. Compared with the prior art, the invention can flexibly carry out frequency spectrum shaping according to the requirements, and realize customized frequency spectrum generation, thereby expanding the application range of the optical frequency comb. In addition, by constructing a reward function and designing action execution and reward feedback of an agent and an experimental environment, the invention effectively optimizes the training process of a phase modulation decision and realizes spectrum programming and control of an optical frequency comb. Compared with the prior art, the deep reinforcement learning algorithm adopted by the invention can more accurately adjust the frequency spectrum parameters, so that the performance parameters of the optical frequency comb are still stable in a complex environment, and the robustness and reliability of the system are improved.
The foregoing embodiments may be partially modified in numerous ways by those skilled in the art without departing from the principles and spirit of the invention, the scope of which is defined in the claims and not by the foregoing embodiments, and all such implementations are within the scope of the invention.
Claims (7)
1. A spectrum programmable optical frequency comb generating method based on deep reinforcement learning is characterized in that an intelligent body based on the deep reinforcement learning is deployed on an experimental physical structure of a broadband optical frequency comb, and an interaction model between the intelligent body and an experimental environment is constructed; constructing a strategy algorithm framework based on a deep reinforcement learning Actor-Critic architecture, and designing interaction content and rules between an intelligent agent and an experimental environment according to the strategy algorithm framework; and constructing a reward function by taking root mean square error of a target frequency spectrum and an experimental frequency spectrum as parameters, designing action execution and reward feedback of an agent and an experimental environment, and obtaining an optimal phase modulation decision by a training strategy of a deep reinforcement learning algorithm so as to realize the programmable generation of an optical frequency comb by the spectrum.
2. The method for generating the spectrum programmable optical frequency comb based on the deep reinforcement learning according to claim 1, which is characterized by comprising the following steps:
step A, constructing a stable electro-optic optical frequency comb, regulating and controlling spectrum phase, and widening the spectrum of the optical frequency comb by utilizing nonlinear effect, wherein the method specifically comprises the following steps of:
a1, generating a stable initial electro-optic frequency comb by a single wavelength laser, one intensity modulator and two phase modulators in cascade, wherein: the single-wavelength laser outputs input light with the wavelength of 1550nm, the power of the input light is 10dBm, the center frequency of the generated initial optical frequency comb is 193.548THz, and the repetition rate is 10GHz;
a2, modulating the phase of the initial optical frequency comb by using a programmable optical processor: selecting a phase within 8nm bandwidth near the center frequency of an initial optical frequency comb, namely 1546nm to 1554nm, as a modulation object, wherein a phase modulation curve is represented by a weighted sum of 20-order chebyshev polynomials with random weights, specifically: chebyshev polynomial function T with nth power n (x)＝cos(ncos -1 (x))，Wherein: w (w) k Representing chebyshev polynomials T k (x) Weight, W of (2) 20 (x) Representing the result of weighted addition of chebyshev polynomials;
a3, inputting the modulated electro-optical frequency comb into an erbium-doped optical fiber amplifier, injecting into a high-nonlinearity optical fiber, and utilizing a nonlinearity effect to perform optical frequency comb spectrum broadening;
a4, acquiring and expanding spectrum information by using a spectrometer, and analyzing to obtain a spectrum initial state and a spectrum execution state;
and B, based on the physical structure of the broadband optical frequency comb obtained through experiments, namely the spectrum initial state and the spectrum execution state obtained in the step A, deploying the deep reinforcement learning agent, establishing a strategy algorithm framework based on a deep reinforcement learning Actor-Critic framework, and designing interaction content and rules of the deep reinforcement learning agent and an experimental environment module, wherein the method specifically comprises the following steps:
b1, constructing a deep reinforcement learning agent module based on an Actor-Critic architecture, wherein: the Actor network generates an action strategy, such as outputting a phase modulation decision, and updates the action strategy according to a cost function Q feedback provided by the Critic network so as to improve the performance of the intelligent agent in the environment; the Critic network evaluates the merits of the action strategy and calculates a cost function Q according to the action taken by the agent so as to evaluate the merits of the current action strategy, and stores the state conversion process after the Actor network interacts with the environment in an experience playback memory;
b2, training the intelligent agent through a network random sampling experience playback memory, and specifically comprises the following steps:
(1) setting a reward function R in view of the Root Mean Square Error (RMSE) of the target spectrum and the experimental spectrum being a key indicator for assessing the control and shaping effects of the optical comb t ＝-RMSE(S target ，S exp ) Wherein: target frequency spectrum S target Experiment frequency spectrum S exp The negative sign is to convert RMSE into a maximization problem, and the task of the agent becomes to minimize the difference between the target spectrum and the experimental spectrum;
(2) the goal of setting the agent is to learn an optimal policy function pi (a t |s t θ), wherein: a, a t Action s generated for agent through Actor network t The state of the experimental environment is that theta is a parameter of an Actor network; policy gradients in an Actor networkWherein: a(s) t ,a t ) Is a dominance function of the Critic network; calculating a loss function->Wherein: e represents an expected value, KL represents KL divergence, θ old Representing old Actor network parameters, lambda is a super parameter for controlling the balance of policy update and policy stability;
(3) evaluating the merits of action strategies generated by an Actor network through a Critic network: setting a cost function Q(s) of a Critic network t ,a t ) I.e. in state s t Take action a t Updating using a method based on TD error with the dominant function: delta t ＝r t +γQ(s t+1 ,a t+1 )-Q(s t ,a t ) Dominance function A (s t ,a t )＝Q(s t ,a t )-V(s t ) Wherein: v(s) t ) As a state value function of the Critic network, reducing variance and improving convergence rate of an algorithm by adopting an advantage function;
b3, adding a small Gaussian noise to the output of the Actor network in each time step to encourage the agent to explore a new action strategy and avoid over-dependence on past experience; specifically, for the generated action a t The added gaussian noise obeys a gaussian distribution with zero mean and sigma standard deviation: a' t ＝a t +n (0, σ), wherein: a, a t Representing the original action generated by the agent at time step t, i.e., the phase modulation decision, which is the output of the Actor network; a' t Representing an action after adding gaussian noise, which is to be performed in a real environment; n (0, σ) represents a gaussian distribution with zero mean and σ standard deviation.
3. The method for generating a spectrum programmable optical frequency comb based on deep reinforcement learning according to claim 2, wherein the intensity modulator is loaded with a direct-current bias voltage to make the direct-current bias voltage work at a positive intersection point, so as to generate an initial optical frequency comb with flat frequency spectrum and more comb teeth.
4. The method for generating a spectrum programmable optical frequency comb based on deep reinforcement learning as claimed in claim 2, wherein the modulated initial optical frequency comb is amplified to 23dBm and then injected into a highly nonlinear optical fiber with a length of 2km for nonlinear broadening, and the nonlinear parameter is 0.03 ps/(nm) 2 ·km)。
5. The method for generating a deep reinforcement learning-based optical frequency comb of claim 2, wherein updating the Critic network uses a mean square error as a loss function:
6. the method for generating spectral programmable optical frequency comb based on deep reinforcement learning as set forth in claim 2, wherein said agent model is trained to obtain an optimal strategy function pi (a by rewarding function and interaction with experimental environment t |s t θ), the initial optical frequency comb after modulation will have a target spectrum defined by user programming through the stretched optical frequency comb obtained by the high nonlinear fiber according to the optimal phase control decision of the agent output.
7. A spectrally programmable optical frequency comb generation system implementing the depth reinforcement learning based spectral programmable optical frequency comb generation method of any one of claims 1-6, comprising: the laser, polarization controller, intensity modulator, phase modulator, programmable light processor, ooze bait fiber amplifier, nonlinear optical fiber and spectrum appearance that link to each other in proper order, wherein: the intensity modulator and the phase modulator control modulation parameters through the radio frequency source and the phase shifter, the spectrometer generates a spectrum initial state and a spectrum execution state according to the acquired broadened spectrum information, and the intelligent agent is used for controlling the phase of the programmable light processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311082437.XA CN117192865A (en) | 2023-08-28 | 2023-08-28 | Spectrum programmable optical frequency comb generation method based on deep reinforcement learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311082437.XA CN117192865A (en) | 2023-08-28 | 2023-08-28 | Spectrum programmable optical frequency comb generation method based on deep reinforcement learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117192865A true CN117192865A (en) | 2023-12-08 |
Family
ID=88998948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311082437.XA Pending CN117192865A (en) | 2023-08-28 | 2023-08-28 | Spectrum programmable optical frequency comb generation method based on deep reinforcement learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117192865A (en) |
-
2023
- 2023-08-28 CN CN202311082437.XA patent/CN117192865A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiang et al. | Fiber laser development enabled by machine learning: review and prospect | |
| Liu et al. | Analytic study on interactions between periodic solitons with controllable parameters | |
| Larger et al. | High-speed photonic reservoir computing using a time-delay-based architecture: Million words per second classification | |
| Genty et al. | Machine learning and applications in ultrafast photonics | |
| Duport et al. | Fully analogue photonic reservoir computer | |
| Fiers et al. | Nanophotonic reservoir computing with photonic crystal cavities to generate periodic patterns | |
| Duport et al. | All-optical reservoir computing | |
| Lucas et al. | Tailoring microcombs with inverse-designed, meta-dispersion microresonators | |
| Yan et al. | Low-latency deep-reinforcement learning algorithm for ultrafast fiber lasers | |
| Harkhoe et al. | Delay-based reservoir computing using multimode semiconductor lasers: Exploiting the rich carrier dynamics | |
| Qin et al. | Numerical simulation and experiment on optical packet header recognition utilizing reservoir computing based on optoelectronic feedback | |
| Nikolić et al. | Breathers, solitons and rogue waves of the quintic nonlinear Schrödinger equation on various backgrounds | |
| Antonik et al. | Online training for high-performance analogue readout layers in photonic reservoir computers | |
| Schneider et al. | Using digital masks to enhance the bandwidth tolerance and improve the performance of on-chip reservoir computing systems | |
| CN106848825A (en) | By cascading the method that optical modulator produces super flat frequency comb | |
| Hou et al. | Parallel information processing using a reservoir computing system based on mutually coupled semiconductor lasers | |
| Zhu et al. | Predicting the dynamic process and model parameters of vector optical solitons under coupled higher-order effects via WL-tsPINN | |
| Zang et al. | Principle-driven fiber transmission model based on PINN neural network | |
| CN106802498B (en) | A kind of all-digital programmable optical frequency comb generation method and device | |
| Akrout et al. | Parallel photonic reservoir computing using frequency multiplexing of neurons | |
| Liu et al. | Adaptive photonic microwave instantaneous frequency estimation using machine learning | |
| Botía et al. | Fuzzy cellular automata and intuitionistic fuzzy sets applied to an optical frequency comb spectral shape | |
| Ma et al. | Artificial intelligence-enabled mode-locked fiber laser: a review | |
| CN117192865A (en) | Spectrum programmable optical frequency comb generation method based on deep reinforcement learning | |
| Fang et al. | Data-driven prediction of soliton solutions of the higher-order NLSE via the strongly-constrained PINN method |
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 |