论文标题
使用强大的元学习方法的选择性分类
Selective classification using a robust meta-learning approach
论文作者
论文摘要
预测不确定性-A模型对通过培训干预措施以及测试时间应用（例如选择性分类）构建强大模型的输入密钥的准确性的自我意识。我们提出了一种新颖的实例条件重新加权方法，该方法使用辅助网络捕获预测性不确定性，并统一这些火车和测试时间应用程序。辅助网络在二聚体优化框架中使用元视线进行训练。我们的提案的关键贡献是将辍学差异（贝叶斯预测不确定性的近似值）最小化的元观点。我们在受控的实验中显示，我们通过这种元观察者有效地捕获了不同特定的不确定性概念，而先前的方法仅捕获某些方面。这些结果转化为现实环境选择性分类，标签噪声，适应性，校准以及跨数据集 - imagenet，cifar100，糖尿病性视网膜病变，Camelyon，camelyon，wild，imagenet-c，-a，-a，-r，clothing1m，clothing1m overinopation offication and codienopation and of 3.4％％/3％的3.4％/3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的3.4％的人选择性分类。我们还改进了大规模预处理的模型，例如PLEX。
Predictive uncertainty-a model's self awareness regarding its accuracy on an input-is key for both building robust models via training interventions and for test-time applications such as selective classification. We propose a novel instance-conditioned reweighting approach that captures predictive uncertainty using an auxiliary network and unifies these train- and test-time applications. The auxiliary network is trained using a meta-objective in a bilevel optimization framework. A key contribution of our proposal is the meta-objective of minimizing the dropout variance, an approximation of Bayesian Predictive uncertainty. We show in controlled experiments that we effectively capture the diverse specific notions of uncertainty through this meta-objective, while previous approaches only capture certain aspects. These results translate to significant gains in real-world settings-selective classification, label noise, domain adaptation, calibration-and across datasets-Imagenet, Cifar100, diabetic retinopathy, Camelyon, WILDs, Imagenet-C,-A,-R, Clothing1M, etc. For Diabetic Retinopathy, we see upto 3.4%/3.3% accuracy and AUC gains over SOTA in selective classification. We also improve upon large-scale pretrained models such as PLEX.
