Multi-view Metric Learning in Vector-valued Kernel Spaces

Riikka Huusari (1, 2, 3) , Hachem Kadri (1, 2, 3) , Cécile Capponi (1, 2, 3)

1 LIS - Laboratoire d'Informatique et des Systèmes (LIS) (Marseille, Toulon)
2 AMU - Aix Marseille Université
3 QARMA - éQuipe d'AppRentissage de MArseille

We consider the problem of metric learning for multi-view data and present a novel method for learning within-view as well as between-view metrics in vector-valued kernel spaces, as a way to capture multi-modal structure of the data. We formulate two convex optimization problems to jointly learn the metric and the classifier or regressor in kernel feature spaces. An iterative three-step multi-view metric learning algorithm is derived from the optimization problems. In order to scale the computation to large training sets, a block-wise Nyström approximation of the multi-view kernel matrix is introduced. We justify our approach theoretically and experimentally, and show its performance on real-world datasets against relevant state-of-the-art methods.

Domaines: Apprentissage [cs.LG]; Machine Learning [stat.ML]

Riikka Huusari, Hachem Kadri, Cécile Capponi. Multi-view Metric Learning in Vector-valued Kernel Spaces. The 21st International Conference on Artificial Intelligence and Statistics, Apr 2018, Lanzarote, Spain. ⟨hal-01736068⟩