3 papers
Revisiting Thinning Methods for Kernel Learning Problems
Blanca Cano-Camarero, Yago R. Aguado-Carrillo-de-Albornoz, Ángela Fernández-Pascual +1
Kernel methods are widely used because of their strong theoretical guarantees and empirical performance. However, their high computational cost limits their applicability to large-…
Contrastive-Collapsed Loss for Flexible and Geometrically Optimal Embeddings and Faster Convergence
Blanca Cano-Camarero, Ángela Fernández-Pascual, José R. Dorronsoro
In this work, we introduce CoCo, a loss function aimed at learning normalized and well-structured representations. The proposed loss encourages intra-class collapse and inter-class…
Faster SVM Training via Conjugate SMO
Alberto Torres-Barrán, Carlos Alaíz, José R. Dorronsoro
We propose an improved version of the SMO algorithm for training classification and regression SVMs, based on a Conjugate Descent procedure. This new approach only involves a modes…