Affinity guided Geometric Semi-Supervised Metric Learning
This work addresses a gap in deep metric learning for semi-supervised scenarios, offering a novel approach that could benefit applications requiring limited labeled data.
The paper tackles the understudied problem of deep semi-supervised distance metric learning by introducing a method that uses Riemannian optimization and affinity propagation for triplet mining, achieving superior performance over competitors.
In this paper, we revamp the forgotten classical Semi-Supervised Distance Metric Learning (SSDML) problem from a Riemannian geometric lens, to leverage stochastic optimization within a end-to-end deep framework. The motivation comes from the fact that apart from a few classical SSDML approaches learning a linear Mahalanobis metric, deep SSDML has not been studied. We first extend existing SSDML methods to their deep counterparts and then propose a new method to overcome their limitations. Due to the nature of constraints on our metric parameters, we leverage Riemannian optimization. Our deep SSDML method with a novel affinity propagation based triplet mining strategy outperforms its competitors.