Multitarget Robust Deep Stochastic Configuration Network Parameter Modeling Method

To improve the model accuracy of deep stochastic configuration network (DSCN) in multitarget robust parameter modeling tasks, this paper presents a multitarget robust DSCN modeling method. This method expands the hidden layer output matrix by optimizing the network model structure and exploiting the correlation between multiple targets to improve the modeling quality. The model output weights are optimized by constructing a sparse weighted robust loss function that combines the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$L2$</tex>,1-norm with the Laplace kernel function, which reduces the model's sensitivity to abnormal training data. Experimental tests are performed on multitarget benchmark datasets to validate the modeling performance of the method. The experimental results show that the multitarget parameter model constructed via the proposed method has advantages in terms of accuracy and robustness.