Tensor-Based Type-2 Extreme Learning Machine with L1-norm and Liu Regression
Jie Li, Guoliang Zhao · 2022 41st Chinese Control Conference (CCC) · 2022
In the paper, tensor-based type-2 extreme learning single layer neural network is extended with L1-norm and Liu regression. The L1-norm and Liu regression regularization methods are merged into tensor-based fuzzy learning system. The antecedent part of the tensor-based fuzzy learning system follows the structure of the tensor-based type-2 extreme learning single layer neural network. To the consequent part of the improved tensor-based type-2 extreme learning single layer neural network, it is a regression problem, the matrix equation that is unfolded from generated tensor is used to solve the learning problem in the consequent part, which provides a basis for the introduction of regularization concurrently. The regularization method is similarly constructed as elastic-net-ELM, the Lasso regression and Liu regression are applied to tensor-based type-2 extreme learning single layer neural network, and L1-norm and L2-norm are used to enhance the tensor-based type-2 extreme learning single layer neural network. To the improved tensor-based type-2 extreme learning single layer neural network, the segment of variable selection of elastic-net-ELM is inherited by the tensor-based type-2 extreme learning single layer neural network with Lasso regression and Liu regression, the system performance is improved further by adjusting the two parameters of the tensor-based type-2 extreme learning single layer neural network.