arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2402.08182cs.LGstat.ML
收藏

变分持续测试时间适应

Variational Continual Test-Time Adaptation

Computer Vision Center (CVC), Universitat Autònoma de Barcelona(计算机视觉中心(CVC),巴塞罗那自治大学);New Laboratory of Pattern Recognition, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences(模式识别新实验室,多模态人工智能系统国家重点实验室,中国科学院自动化研究所);School of Automation, Southeast University(自动化学院,东南大学);China Nuclear Power Engineering Co.,LTD(中国核电工程有限公司);School of Electronic & Engineering, Suzhou University of Science and Technology(电子与工程学院,苏州科技大学)

Fan Lyu, Kaile Du, Yuyang Li, Hanyu Zhao, Fuyuan Hu, Zhang Zhang, Guangcan Liu, Liang Wang

AI总结:

本文提出VCoTTA方法,通过变分贝叶斯方法解决持续测试时间适应中的不确定性问题,结合源和教师模型的先验信息以缓解误差累积。

摘要:

持续测试时间适应(CTTA)任务研究了在测试时间内连续领域偏移情况下有效的领域适应。由于仅使用未标记样本,模型更新存在显著不确定性,导致CTTA面临严重的误差累积问题。在本文中,我们引入了VCoTTA,一种变分贝叶斯方法,用于测量CTTA中的不确定性。在源阶段,我们通过变分预热策略将预训练的确定性模型转换为贝叶斯神经网络(BNN),向模型注入不确定性。在测试时间,我们使用变分推断对学模型和指数移动平均对教师模型进行均教更新策略。我们的新方法通过结合源和教师模型的先验信息来更新学模型。证据下界被公式化为学模型和教师模型之间的交叉熵,以及先验混合的KL散度。在三个数据集上的实验结果表明,该方法在CTTA框架中有效缓解了误差累积问题。

英文摘要:

Continual Test-Time Adaptation (CTTA) task investigates effective domain adaptation under the scenario of continuous domain shifts during testing time. Due to the utilization of solely unlabeled samples, there exists significant uncertainty in model updates, leading CTTA to encounter severe error accumulation issues. In this paper, we introduce VCoTTA, a variational Bayesian approach to measure uncertainties in CTTA. At the source stage, we transform a pretrained deterministic model into a Bayesian Neural Network (BNN) via a variational warm-up strategy, injecting uncertainties into the model. During the testing time, we employ a mean-teacher update strategy using variational inference for the student model and exponential moving average for the teacher model. Our novel approach updates the student model by combining priors from both the source and teacher models. The evidence lower bound is formulated as the cross-entropy between the student and teacher models, along with the Kullback-Leibler (KL) divergence of the prior mixture. Experimental results on three datasets demonstrate the method's effectiveness in mitigating error accumulation within the CTTA framework.

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。

↑