On Episodes, Prototypical Networks, and Few-shot Learning

By Steinar Laenen, Luca Bertinetto. 2020

136 citations recorded by the source corpus.

Episodic learning is a popular practice among researchers and practitioners interested in few-shot learning. It consists of organising training in a series of learning problems, each relying on small "support" and "query" sets to mimic the few-shot circumstances encountered during evaluation. In this paper, we investigate the usefulness of episodic learning in Prototypical Networks and Matching Networks, two of the most popular algorithms making use of this practice. Surprisingly, in our experiments we found that, for Prototypical and Matching Networks, it is detrimental to use the episodic learning strategy of separating training samples between support and query set, as it is a data-inefficient way to exploit training batches. These "non-episodic" variants, which are closely related to the classic Neighbourhood Component Analysis, reliably improve over their episodic counterparts in multiple datasets, achieving an accuracy that (in the case of Prototypical Networks) is competitive with the state-of-the-art, despite being extremely simple.

Subfields: Meta & Continual Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Meta & Continual Learning

1. Katja Hofmann
2. Luisa M. Zintgraf
3. Daniela Massiceti
4. John F. Bronskill
5. Chelsea Finn
6. Sergey Levine

Citations in the corpus

• ORBIT: A Real-World Few-Shot Dataset for Teachable Object Recognition (2021, methodology): “The recognition model can be trained on these tasks using an episodic [9, 40, 49, 38] or non-episodic approach [5, 46, 22].”