arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.14707cs.LGstat.ML
收藏

基于合成数据的上下文无关学习

Towards Context-Agnostic Learning Using Synthetic Data

CSAIL(麻省理工学院计算机科学与人工智能实验室);MIT(麻省理工学院)

Charles Jin, Martin Rinard

摘要:

我们提出了一种新的学习设置,其中输入域是定义在两个集合的乘积上的映射的图像,其中一个集合完全确定标签。我们推导出一个新的风险约束,分解成一个偏差和一个错误项,并表现出令人惊讶的弱依赖于真实的标签。受这些结果的启发,我们提出了一种算法,旨在最大限度地减少偏差项利用的能力,从每一个独立的集的样本。我们将我们的设置应用于视觉分类任务,我们的方法使我们能够在完全由每个类的单个合成示例组成的数据集上训练分类器。在现实世界的图像分类的几个标准基准上,我们在上下文不可知的设置中实现了鲁棒的性能,具有对现实世界域的良好概括,而直接在现实世界数据上训练,而不使用我们的技术,则会产生对背景扰动很脆弱的分类器。

英文摘要:

We propose a novel setting for learning, where the input domain is the image of a map defined on the product of two sets, one of which completely determines the labels. We derive a new risk bound for this setting that decomposes into a bias and an error term, and exhibits a surprisingly weak dependence on the true labels. Inspired by these results, we present an algorithm aimed at minimizing the bias term by exploiting the ability to sample from each set independently. We apply our setting to visual classification tasks, where our approach enables us to train classifiers on datasets that consist entirely of a single synthetic example of each class. On several standard benchmarks for real-world image classification, we achieve robust performance in the context-agnostic setting, with good generalization to real world domains, whereas training directly on real world data without our techniques yields classifiers that are brittle to perturbations of the background.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑