Prototypical Networks for Few-shot Learning
Jake Snell University of Toronto111Initial work done while at Twitter.、Kevin Swersky Twitter、Richard S. Zemel University of Toronto, Vector Institute
Abstract
We propose prototypical networks for the problem of few-shot classification, where a classifier must generalize to new classes not seen in the training set, given only a small number of examples of each new class. Prototypical networks learn a metric space in which classification can be performed by computing distances to prototype representations of each class. Compared to recent approaches for few-shot learning, they reflect a simpler inductive bias that is beneficial in this limited-data regime, and achieve excellent results. We provide an analysis showing that some simple design decisions can yield substantial improvements over recent approaches involving complicated architectural choices and meta-learning. We further extend prototypical networks to zero-shot learning and achieve state-of-the-art results on the CU-Birds dataset.
中文速览
小样本分类（few-shot classification）要求模型仅凭每个新类别的极少量样本就能正确识别，传统深度学习方法在这种数据极度匮乏的场景下极易过拟合。原型网络（Prototypical Networks）的思路非常直接：用神经网络把所有样本映射到一个嵌入空间，然后把每个类别的支撑样本取平均得到该类的"原型"向量，对新样本的分类就变成"找最近的原型"这一简单的距离计算。研究发现，使用欧氏距离（而非常见的余弦相似度）、以及在训练时采用比测试时更多类别数的回合设置，能带来显著的性能提升，并且在 Omniglot 和 miniImageNet 两个标准基准上达到了当时最优水平，同时还将该方法扩展到零样本学习（zero-shot learning）并在 CUB-200 鸟类数据集上取得最佳结果。这项工作的意义在于，它用一个理论上有据可依（等价于用 Bregman 散度做混合密度估计）、工程上极为简洁的方案，超越了依赖复杂元学习机制的方法，为小样本学习提供了一个强力且易于复现的基线。
原文 arXiv:1703.05175;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1703.05175v2