Are Few-Shot Learning Benchmarks too Simple ? Solving them without Task Supervision at Test-Time
This work addresses the inadequacy of current benchmarks for evaluating few-shot learning methods, which is a problem for researchers in machine learning, though it is incremental as it builds on existing methods like Prototypical Networks.
The authors tackled the problem of few-shot learning benchmarks being too simple by showing that several popular benchmarks can be solved without using support set labels at test-time, using a new baseline called Centroid Networks that achieves varying success by clustering to recover hidden labels.
We show that several popular few-shot learning benchmarks can be solved with varying degrees of success without using support set Labels at Test-time (LT). To this end, we introduce a new baseline called Centroid Networks, a modification of Prototypical Networks in which the support set labels are hidden from the method at test-time and have to be recovered through clustering. A benchmark that can be solved perfectly without LT does not require proper task adaptation and is therefore inadequate for evaluating few-shot methods. In practice, most benchmarks cannot be solved perfectly without LT, but running our baseline on any new combinations of architectures and datasets gives insights on the baseline performance to be expected from leveraging a good representation, before any adaptation to the test-time labels.