Variable Selection in Distance Metric Learning and Triplet Constraints for Deep Learning Based Ordinal Classification

The purpose of this research is to augment linear and kernelized ordinal distance metric learning (L/KODML) techniques with a proposed variable selection methodology that integrates the Sequential Multi-Response Feature Selection (SMuRFS) algorithm. Additionally, we aim to embed ordinal triplet constraints into a deep learning architecture, and to propose a general framework for deep learning-based ordinal classification. A variety of simulation studies and real data experiments were conducted to evaluate the various methodologies. For the distance metric learning and variable selection, results showed that the integration of SMuRFS performed effective variable selection and improved prediction accuracy. For the triplet constraints, incorporating ordinal triplets into a deep learning framework yielded improved classification results. Key findings imply that SMuRFS can be extended to incorporate different predictive mechanisms, and that utilizing ordinal triplets in a deep learner is a viable approach for ordinal classification. These principal outcomes open the door for various avenues of future research, such as the choice of transformation and influence functions within the SMuRFS framework. Future research may also explore triplet mining techniques and the integration of ordinal triplets into CNN-based architectures. Advisor: Souparno Ghosh