一句话：提出多边界聚类与优先级排序，从测试输入中选取有效子集，以有限标注促进神经网络重训练。
深度学习模型在安全攸关场景中广泛应用，但训练数据与测试数据之间存在数据鸿沟，导致模型在测试上下文中的表现难以保证。为以有限标注代价提升模型质量，论文提出多边界聚类与优先级排序方法，从测试输入中挑选有效子集用于神经网络重训练。其核心思路是利用多个边界信息对测试样本进行聚类，并对样本优先级排序，从而在有限标注预算下选出对重训练最有价值的输入。摘要未报告具体实验指标，但该工作面向 DL 测试中测试输入选择与重训练数据增强这一关键问题。
With the increasing application of deep learning (DL) models in many safety-critical scenarios, effective and efficient DL testing techniques are much in demand to improve the quality of DL models. One of the major challenges is the data gap between the training data to construct the models and the testing data to evaluate them. To bridge the gap, testers aim to collect an effective subset of inputs from the testing contexts, with limited labeling effort, for retraining DL models.
DOI 原文 ·
@inproceedings{paperbot6012,
title = {Multiple-boundary clustering and prioritization to promote neural network retraining},
author = {Weijun Shen and Yanhui Li and Lin Chen and Yuanlei Han and Yuming Zhou and Baowen Xu},
booktitle = {ASE},
year = {2020},
doi = {10.1145/3324884.3416621}
}