arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2401.12129cs.CVcs.LG
收藏

基于消融学习温度能量的分布外检测及应用

Out-of-Distribution Detection & Applications With Ablated Learned Temperature Energy

Cranberry-Lemon University(蔓越莓柠檬大学)

Will LeVine, Benjamin Pikus, Jacob Phillips, Berk Norman, Fernando Amat Gil, Sean Hendryx

AI总结:

本文提出AbeT方法,通过消融学习温度能量实现分布外检测,在分类任务FPR@95降低43.43%,并在目标检测和语义分割中提升性能,仅需单阶段训练且无额外超参数。

摘要:

随着深度神经网络在高风险领域的应用,识别推理输入何时属于分布外(OOD)至关重要,这样用户就能在高置信度下被提醒性能和校准可能下降——最终了解何时应信任网络的决策(及其决策中的不确定性)。本文提出消融学习温度能量(简称“AbeT”),这是一种OOD检测方法,在分类任务中,与现有技术相比,在95%真阳性率下的假阳性率(FPR@95)降低了43.43%,且无需多阶段训练网络、无需超参数或测试时反向传播。我们还提供了经验性见解,解释为何我们的模型仅通过在训练时接触错误分类的分布内(ID)样本,就能学会区分ID和OOD样本。最后,我们展示了该方法在目标检测和语义分割中分别识别OOD对象对应的预测边界框和像素的有效性:与先前的现有技术相比,目标检测的AUROC提高了5.15%,语义分割的FPR@95降低了41.48%且AUPRC提高了34.20%。

英文摘要:

As deep neural networks become adopted in high-stakes domains, it is crucial to identify when inference inputs are Out-of-Distribution (OOD) so that users can be alerted of likely drops in performance and calibration despite high confidence -- ultimately to know when networks' decisions (and their uncertainty in those decisions) should be trusted. In this paper we introduce Ablated Learned Temperature Energy (or "AbeT" for short), an OOD detection method which lowers the False Positive Rate at 95\% True Positive Rate (FPR@95) by $43.43\%$ in classification compared to state of the art without training networks in multiple stages or requiring hyperparameters or test-time backward passes. We additionally provide empirical insights as to why our model learns to distinguish between In-Distribution (ID) and OOD samples while only being explicitly trained on ID samples via exposure to misclassified ID examples at training time. Lastly, we show the efficacy of our method in identifying predicted bounding boxes and pixels corresponding to OOD objects in object detection and semantic segmentation, respectively -- with an AUROC increase of $5.15\%$ in object detection and both a decrease in FPR@95 of $41.48\%$ and an increase in AUPRC of $34.20\%$ in semantic segmentation compared to previous state of the art.

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。

↑