Optical reservoir computing based on the bagging trees

In traditional optical reservoir computing, the least squares method are commonly used to train the output weights for regression tasks. Although these algorithms are highly versatile, their training efficiency and accuracy are somewhat inferior compared to current algorithms. Bagging trees, an ensemble learning method, works by resampling the dataset to train multiple models and then combining the predictions of these models, thus improving the stability and accuracy of the final prediction. By combining optical reservoir computing with the bagging trees, both the prediction accuracy and training efficiency are greatly improved, with the highest R-squared prediction reaching 99.56.