Doubly Regularized Least Squares Twin Extreme Learning Machine for Pattern Classification
Qing Wu, Ming Yu, Lu Chen · 2023
Least squares twin extreme learning machine (LSTELM) possesses better generalization ability than the standard extreme learning machine (ELM) by finding two nonparallel hyperplanes in ELM random feature space. However, LSTELM is not sparse in nature and does not take into account the intrinsic geometric information of the data. To overcome the above, this paper proposes one classification method based on LSTELM, namely doubly regularized least squares twin extreme learning machine (DRLSTELM). DRLSTELM uses L2,1-norm regularization and manifold regularization methods to improve the classification performance. Furthermore, in its nonlinear case, a hybrid kernel function is introduced to improve the generalization and learning ability of the model. To evaluate the classification performance of the proposed algorithm, experiments were conducted on the NDC datasets and the benchmark datasets respectively. The results show that the classification effect of DRLSTELM is significantly better than other state-of-the-art methods.