研究目的
To compare the performance of several machine learning algorithms for forecasting the amplitude of upcoming emitted chaotic pulses from an optically injected semiconductor laser.
研究成果
Accurate prediction of the amplitude of upcoming chaotic pulses is possible using machine learning techniques, although the presence of extreme events and stochastic contributions limit the accuracy. The deep neural network, k-nearest neighbors, and reservoir computing methods show the best performance. The study suggests that similar methods may be used in the forecast of more complex systems, but further testing is necessary.
研究不足
The presence of extreme events in the time series and stochastic contributions in the laser model bound the accuracy that can be achieved. The performance of the forecasting methods is also limited by the amount of noise and the length of the time series used for training.
1：Experimental Design and Method Selection:
The study uses a laser model to simulate the dynamics of an optically injected semiconductor laser, focusing on a regime that shows ultrahigh intensity pulses. Machine learning algorithms (deep learning, support vector machine, nearest neighbors, and reservoir computing) are compared for forecasting the amplitude of the next pulse.
2：Sample Selection and Data Sources:
Simulated time series of the laser's output intensity under different dynamical regimes, with and without extreme pulses, are used.
3：List of Experimental Equipment and Materials:
A semiconductor laser model is simulated using rate equations for the complex optical field and carrier population.
4：Experimental Procedures and Operational Workflow:
The Runge-Kutta method of order 2 is used for simulation with a specific time step. The performance of each machine learning algorithm is evaluated based on the mean absolute relative error (MARE).
5：Data Analysis Methods:
The MARE is used to quantify the performance of the forecasting algorithms. The influence of noise and the length of the training dataset on prediction accuracy are analyzed.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
