MIMLTWSVM: Twin support vector machine for multi-instance multi-label learning
Divya Tomar, Sonali Agarwal · 2016
Recently, Multi Instance Multi Label (MIML) learning has attracted the attention of researchers in which an example not only belongs to multiple instances but also associated with multiple class labels. This study proposes a novel multi-instance multi-label twin support vector machine (MIMLTWSVM) classifier by extending the recently proposed binary twin support vector machine (TWSVM) classifier. MIMLTWSVM classifier involves two steps-(1) the problem of multi-instance multi-label has been converted to single instance multi label learning in the first step and (2) in the second step, the derived problem is solved by using multi-label twin support vector machine classifier. The involved Quadratic Programming Problems (QPPs) of proposed classifier has been solved by Successive Over-Relaxation (SOR) technique to speed up the training procedure. The experiment has been conducted on two MIML benchmark datasets-Scene and Reuters. The experimental results demonstrate the superiority of the proposed classifier over several existing state-of-the-art MIML classifiers such as MIMLSVM, MIMLRBF, MIMLBOOST, MIML-kNN and M3MIML.