2014/12/31 by Ilja Kuzborskij, Francesco Orabona · 1 citation
Computer Science · Engineering · Mathematics · #Algorithm #Artificial intelligence #Artificial neural network #Class (philosophy) #Combinatorics #Computer science #Convex combination #Convex optimization #Discrete mathematics #Domain Adaptation and Few-Shot Learning #Focus (optics) #Function (biology) #Generalization #Generalization error #Machine Learning and Algorithms #Mathematics #Regular polygon #Set (abstract data type) #Sparse and Compressive Sensing Techniques #Task (project management) #cs.LG
paper · pdf · doi:10.1007/s10994-016-5594-4
arxiv created 2015/07/17 · openalex publication_date 2016/10/17 · arxiv updated 2016/10/19 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
In this work we consider the learning setting where, in addition to the training set, the learner receives a collection of auxiliary hypotheses originating from other tasks. We focus on a broad class of ERM-based linear algorithms that can be instantiated with any non-negative smooth loss function and any strongly convex regularizer. We establish generalization and excess risk bounds, showing that, if the algorithm is fed with a good combination of source hypotheses, generalization happens at the fast rate O(1/m) instead of the usual O(1/√(m)). On the other hand, if the source hypotheses combination is a misfit for the target task, we recover the usual learning rate. As a byproduct of our study, we also prove a new bound on the Rademacher complexity of the smooth loss class under weaker assumptions compared to previous works.