Genetic Programming for Ensemble Learning in Face Recognition

Tian Zhang,Lianbo Ma,Qunfeng Liu,Nan Li,Yang Liu
DOI: https://doi.org/10.1007/978-3-031-09726-3_19
2022-01-01
Abstract:Ensemble learning has recently been explored to achieve a better generalization ability than a single base learner through combining results of multiple base learners. Genetic programming (GP) can be used to design ensemble learning via different strategies. However, the challenge remains to automatically design an ensemble learning model due to complex search space. In this paper, we propose a new automated ensemble learning framework, based on GP for face recognition, called Evolving Genetic Programming Ensemble Learning (EGPEL). Thismethod integrates feature extraction, base learner selection, and learner hyperparameter optimization, into several program trees. To this end, multiple program trees, a base learner set, and a hyperparameter set are developed in EGPEL. Meanwhile, an evolutionary approach to results integration is proposed. The performance of EGPEL is verified on face benchmark datasets of difficulty and compared with a large number of commonly used peer competitors, including state-of-the-art competitors. The results show that EGPEL performs better than most competitive ensemble learning methods.
What problem does this paper attempt to address?