2021/01/01 by Johannes von Oswald, von Oswald, Johannes, Dominic Zhao +12 · 3 citations
Computer Science · Physics and Astronomy · #Domain Adaptation and Few-Shot Learning #Human Pose and Action Recognition #Model Reduction and Neural Networks #cs.LG #cs.NE
paper · pdf · doi:10.48550/arxiv.2110.14402
Published at NeurIPS 2021
arxiv created 2021/10/27 · arxiv updated 2021/10/28
Finding neural network weights that generalize well from small datasets is difficult. A promising approach is to learn a weight initialization such that a small number of weight changes results in low generalization error. We show that this form of meta-learning can be improved by letting the learning algorithm decide which weights to change, i.e., by learning where to learn. We find that patterned sparsity emerges from this process, with the pattern of sparsity varying on a problem-by-problem basis. This selective sparsity results in better generalization and less interference in a range of few-shot and continual learning problems. Moreover, we find that sparse learning also emerges in a more expressive model where learning rates are meta-learned. Our results shed light on an ongoing debate on whether meta-learning can discover adaptable features and suggest that learning by sparse gradient descent is a powerful inductive bias for meta-learning systems.