Constructive multi-output extreme learning machine with application to large tanker motion dynamics identification

In this paper, a novel constructive multi-output extreme learning machine (CM-ELM) is proposed to deal with a large tanker motion dynamics identification. The significant contributions are as follows. (1) Driven by generated tanker dynamics data from the reference model, the CM-ELM method is proposed to identify multi-output dynamic models. (2) The candidate pool for CM-ELM is randomly generated by the ELM strategy, and ranked chunk-by-chunk based on a novel improved multi-response sparse regression (I-MRSR) incorporated with λ weighting. (3) Consequently, the constructive model selection works with fast speed due to chunk-type training process, which also benefits stable hidden node selection and corresponding generalization. (4) Furthermore, output weight update on the final CM-ELM model randomly selected from the elite subset is conducted to enhance the overall performance of the resulting CM-ELM scheme. Finally, the convincing performance of the complete CM-ELM paradigm is verified by simulation studies on not only tanker motion dynamics identification but also benchmark multi-output regressions. Comprehensive comparisons of the CM-ELM with ELM and OP-ELM indicate the remarkable superiority in terms of generalization capability and stable compact structure. Conclusions are steadily drawn that the CM-ELM method is feasibly effective for tanker motion dynamics identification and multi-output regressions.