PLATINUM: Semi-Supervised Model Agnostic Meta-Learning using Submodular Mutual Information
PLATINUM: Semi-Supervised Model Agnostic Meta-Learning using Submodular Mutual Information
复制标题
DOI：
--
复制
发表时间：
2022-01
期刊：
影响因子：
--
通讯作者：
Changbin Li;S. Kothawade;F. Chen;Rishabh K. Iyer
中科院分区：
文献类型：
--
作者：
Changbin Li;S. Kothawade;F. Chen;Rishabh K. Iyer
少射分类（FSC）需要训练模型使用几个（通常一到五个）数据点每个类。元学习已被证明能够通过训练各种其他分类任务来学习FSC的参数化模型。在这项工作中，我们提出了一个半监督模型不可知论元学习框架PLATINUM (semi-suPervised modeL Agnostic meTa-learnIng usiNg sUbmodular Mutual information)，这是一个新的半监督模型不可知论元学习框架，它使用子模块互信息（SMI）函数来提高FSC的性能。PLATINUM在元训练期间利用SMI函数在内环和外环中利用未标记的数据，并为元测试获得更丰富的元学习参数化。我们在两种情况下研究了PLATINUM的性能：1)未标记的数据点与某一集的标记集属于同一类集，以及2)存在不属于标记集的分布外类。我们在miniImageNet、tieredImageNet和few - shot- cifar100数据集上的各种设置上评估了我们的方法。我们的实验表明，PLATINUM优于MAML和半监督方法（如半监督FSC的伪标记），特别是对于每个类的标记样本比例较小。
Few-shot classification (FSC) requires training models using a few (typically one to five) data points per class. Meta learning has proven to be able to learn a parametrized model for FSC by training on various other classification tasks. In this work, we propose PLATINUM (semi-suPervised modeL Agnostic meTa-learnIng usiNg sUbmodular Mutual information), a novel semi-supervised model agnostic meta-learning framework that uses the submodular mutual information (SMI) functions to boost the performance of FSC. PLATINUM leverages unlabeled data in the inner and outer loop using SMI functions during meta-training and obtains richer meta-learned parameterizations for meta-test. We study the performance of PLATINUM in two scenarios - 1) where the unlabeled data points belong to the same set of classes as the labeled set of a certain episode, and 2) where there exist out-of-distribution classes that do not belong to the labeled set. We evaluate our method on various settings on the miniImageNet, tieredImageNet and Fewshot-CIFAR100 datasets. Our experiments show that PLATINUM outperforms MAML and semi-supervised approaches like pseduo-labeling for semi-supervised FSC, especially for small ratio of labeled examples per class.