XB-MAML: Learning Expandable Basis Parameters for Effective Meta-Learning
XB-MAML introduces a novel meta-learning approach that adaptively increases the number of initialized models and refines the initialization points through linear combinations, contributing to more efficient meta-learning.