Meta Networks
Abstract
Neural networks have been successfully applied in applications with a large amount of labeled data. However, the task of rapid generalization on new concepts with small training data while preserving performances on previously learned ones still presents a significant challenge to neural network models. In this work, we introduce a novel meta learning method, Meta Networks (MetaNet), that learns a meta-level knowledge across tasks and shifts its inductive biases via fast parameterization for rapid generalization. When evaluated on Omniglot and Mini-ImageNet benchmarks, our MetaNet models achieve a near human-level performance and outperform the baseline approaches by up to 6% accuracy. We demonstrate several appealing properties of MetaNet relating to generalization and continual learning.
Cite
@article{arxiv.1703.00837,
title = {Meta Networks},
author = {Tsendsuren Munkhdalai and Hong Yu},
journal= {arXiv preprint arXiv:1703.00837},
year = {2017}
}
Comments
Accepted at ICML 2017 - rewrote: the main section; added: MetaNet algorithmic procedure; performed: Mini-ImageNet evaluation