Bound estimation-based safe acceleration for maximum margin of twin spheres machine with pinball loss
作者：
Highlights：
• We estimate center term by variational inequality and utilize v-property to estimate the upper and lower bounds of R and T. The proposed method can safely accelerate a family of hypersphere support vector machine.
• Through our method, the scale of the problem could be greatly reduced, and thus the training time is significantly saved. Moreover, it does not sacrifice the optimal solutions.
• Our method deletes samples from corresponding classes separately. M ore majority samples will be discarded. Therefore, our method is suitable to imbalanced data.
摘要
•We estimate center term by variational inequality and utilize v-property to estimate the upper and lower bounds of R and T. The proposed method can safely accelerate a family of hypersphere support vector machine.•Through our method, the scale of the problem could be greatly reduced, and thus the training time is significantly saved. Moreover, it does not sacrifice the optimal solutions.•Our method deletes samples from corresponding classes separately. M ore majority samples will be discarded. Therefore, our method is suitable to imbalanced data.
论文关键词：Maximum margin,Pinball loss,Imbalanced data,Bound estimation,Upper and lower bounds
论文评审过程：Received 7 March 2020, Revised 25 October 2020, Accepted 22 January 2021, Available online 1 February 2021, Version of Record 6 February 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.107860
