Temperature-Sensitive Optimization of VCSEL L-I Models with Regularization and Activation Functions

In this experiment, due to the low fitting accuracy of the traditional L- I model, a new optimization method for the traditional model is proposed, which is applied to training the L-I model. This method allows the model to achieve higher fitting accuracy under various temperature conditions. To improve the model's performance, after training the model using the measurement data at 20°C, we optimized the model parameters and adopted different activation functions and regularization methods, resulting in a significant improvement in the fitting accuracy of the model. Ultimately, the fitting accuracy of the L-I model was improved from the initial R2 = 0.36 to R2 = 0.93. The results of this study indicate that the combination of various optimization methods not only improves the overall performance of the model but also provides a reference for future similar problems.