Timezone: »

On Episodes, Prototypical Networks, and Few-Shot Learning
Steinar Laenen · Luca Bertinetto

Tue Dec 07 08:30 AM -- 10:00 AM (PST) @

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 (or episodes), each divided into a small training and validation subset to mimic the circumstances encountered during evaluation.But is this always necessary?In this paper, we investigate the usefulness of episodic learning in methods which use nonparametric approaches, such as nearest neighbours, at the level of the episode.For these methods, we not only show how the constraints imposed by episodic learning are not necessary, but that they in fact lead to a data-inefficient way of exploiting training batches.We conduct a wide range of ablative experiments with Matching and Prototypical Networks, two of the most popular methods that use nonparametric approaches at the level of the episode.Their "non-episodic'' counterparts are considerably simpler, have less hyperparameters, and improve their performance in multiple few-shot classification datasets.

Author Information

Steinar Laenen (University of Edinburgh)

PhD Student working on spectral clustering/graph clustering algorithms

Luca Bertinetto (University of Oxford)

Luca Bertinetto is a PhD candidate in the Torr Vision Group at the University of Oxford. The main focus of his doctorate is the problem of agnostic object tracking, which he likes to tackle using simple and effective approaches. Before getting lost among the spires of Oxford, he obtained a joint MSc in Computer Engineering between the Polytechnic University of Turin and Telecom Paris Tech. He has published at CVPR and NIPS and reviewed for PAMI.

More from the Same Authors