Few-shot classification refers to learning a classifier for new classes given
only a few examples. While a plethora of models have emerged to tackle this
recently, we find the current procedure and datasets that are used to
systematically assess progress in this setting lacking. To address this, we
propose Meta-Dataset: a new benchmark for training and evaluating few-shot
classifiers that is large-scale, consists of multiple datasets, and presents
more natural and realistic tasks. The aim is to measure the ability of
state-of-the-art models to leverage diverse sources of data to achieve higher
generalization, and to evaluate that generalization ability in a more
challenging setting. We additionally measure robustness of current methods to
variations in the number of available examples and the number of classes.
Finally our extensive empirical evaluation leads us to identify weaknesses in
Prototypical Networks and MAML, two popular few-shot classification methods,
and to propose a new method, Proto-MAML, which achieves improved performance on
our benchmark.
Description
[1903.03096v1] Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples
%0 Journal Article
%1 triantafillou2019metadataset
%A Triantafillou, Eleni
%A Zhu, Tyler
%A Dumoulin, Vincent
%A Lamblin, Pascal
%A Xu, Kelvin
%A Goroshin, Ross
%A Gelada, Carles
%A Swersky, Kevin
%A Manzagol, Pierre-Antoine
%A Larochelle, Hugo
%D 2019
%K few-shot learning
%T Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few
Examples
%U http://arxiv.org/abs/1903.03096
%X Few-shot classification refers to learning a classifier for new classes given
only a few examples. While a plethora of models have emerged to tackle this
recently, we find the current procedure and datasets that are used to
systematically assess progress in this setting lacking. To address this, we
propose Meta-Dataset: a new benchmark for training and evaluating few-shot
classifiers that is large-scale, consists of multiple datasets, and presents
more natural and realistic tasks. The aim is to measure the ability of
state-of-the-art models to leverage diverse sources of data to achieve higher
generalization, and to evaluate that generalization ability in a more
challenging setting. We additionally measure robustness of current methods to
variations in the number of available examples and the number of classes.
Finally our extensive empirical evaluation leads us to identify weaknesses in
Prototypical Networks and MAML, two popular few-shot classification methods,
and to propose a new method, Proto-MAML, which achieves improved performance on
our benchmark.
@article{triantafillou2019metadataset,
abstract = {Few-shot classification refers to learning a classifier for new classes given
only a few examples. While a plethora of models have emerged to tackle this
recently, we find the current procedure and datasets that are used to
systematically assess progress in this setting lacking. To address this, we
propose Meta-Dataset: a new benchmark for training and evaluating few-shot
classifiers that is large-scale, consists of multiple datasets, and presents
more natural and realistic tasks. The aim is to measure the ability of
state-of-the-art models to leverage diverse sources of data to achieve higher
generalization, and to evaluate that generalization ability in a more
challenging setting. We additionally measure robustness of current methods to
variations in the number of available examples and the number of classes.
Finally our extensive empirical evaluation leads us to identify weaknesses in
Prototypical Networks and MAML, two popular few-shot classification methods,
and to propose a new method, Proto-MAML, which achieves improved performance on
our benchmark.},
added-at = {2019-03-12T13:16:42.000+0100},
author = {Triantafillou, Eleni and Zhu, Tyler and Dumoulin, Vincent and Lamblin, Pascal and Xu, Kelvin and Goroshin, Ross and Gelada, Carles and Swersky, Kevin and Manzagol, Pierre-Antoine and Larochelle, Hugo},
biburl = {https://www.bibsonomy.org/bibtex/2eadd33e3cbda9dacffd0e3546aeb9b02/kirk86},
description = {[1903.03096v1] Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples},
interhash = {388749861dd66df93efe3354f2b405a3},
intrahash = {eadd33e3cbda9dacffd0e3546aeb9b02},
keywords = {few-shot learning},
note = {cite arxiv:1903.03096},
timestamp = {2019-03-12T13:16:42.000+0100},
title = {Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few
Examples},
url = {http://arxiv.org/abs/1903.03096},
year = 2019
}