Contextual Gradient Scaling for Few-Shot Learning
This addresses a specific bottleneck in meta-learning for few-shot classification, offering an incremental improvement to MAML.
The paper tackled the problem of gradient norm imbalance in MAML for few-shot learning, which causes the backbone to learn task-generic features and reduces adaptation performance. They proposed contextual gradient scaling (CxGrad) to scale backbone gradients in a task-wise manner, improving MAML performance by a significant margin in same- and cross-domain classification.
Model-agnostic meta-learning (MAML) is a well-known optimization-based meta-learning algorithm that works well in various computer vision tasks, e.g., few-shot classification. MAML is to learn an initialization so that a model can adapt to a new task in a few steps. However, since the gradient norm of a classifier (head) is much bigger than those of backbone layers, the model focuses on learning the decision boundary of the classifier with similar representations. Furthermore, gradient norms of high-level layers are small than those of the other layers. So, the backbone of MAML usually learns task-generic features, which results in deteriorated adaptation performance in the inner-loop. To resolve or mitigate this problem, we propose contextual gradient scaling (CxGrad), which scales gradient norms of the backbone to facilitate learning task-specific knowledge in the inner-loop. Since the scaling factors are generated from task-conditioned parameters, gradient norms of the backbone can be scaled in a task-wise fashion. Experimental results show that CxGrad effectively encourages the backbone to learn task-specific knowledge in the inner-loop and improves the performance of MAML up to a significant margin in both same- and cross-domain few-shot classification.