Learning a Universal Template for Few-shot Dataset Generalization
This addresses the problem of adapting models to new datasets with limited examples for researchers in meta-learning, though it is incremental in improving parameter efficiency and scalability.
The paper tackles few-shot dataset generalization by proposing a universal template that defines dataset-specialized models with few inferred parameters, achieving state-of-the-art results on the Meta-Dataset benchmark.
Few-shot dataset generalization is a challenging variant of the well-studied few-shot classification problem where a diverse training set of several datasets is given, for the purpose of training an adaptable model that can then learn classes from new datasets using only a few examples. To this end, we propose to utilize the diverse training set to construct a universal template: a partial model that can define a wide array of dataset-specialized models, by plugging in appropriate components. For each new few-shot classification problem, our approach therefore only requires inferring a small number of parameters to insert into the universal template. We design a separate network that produces an initialization of those parameters for each given task, and we then fine-tune its proposed initialization via a few steps of gradient descent. Our approach is more parameter-efficient, scalable and adaptable compared to previous methods, and achieves the state-of-the-art on the challenging Meta-Dataset benchmark.