arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.05140cs.LGcs.AI
收藏

用于预测性流失的局部自适应标签平滑

Locally Adaptive Label Smoothing for Predictive Churn

Google Research(谷歌研究院)

Dara Bahri, Heinrich Jiang

摘要:

训练现代神经网络是一个固有的噪声过程,可能导致高\emph{prediction churn} -由于参数初始化和小批量中的随机化等因素,同一模型的重新训练之间存在差异-即使训练的模型都达到相似的精度。这种预测扰动在实践中可能是非常不期望的。在本文中,我们提出了几个基线,以减少流失,并表明,软标签的训练,通过自适应平滑每个例子的标签的基础上的例子的相邻标签,往往优于流失的基线,同时提高各种基准分类任务和模型架构的准确性。

英文摘要:

Training modern neural networks is an inherently noisy process that can lead to high \emph{prediction churn} -- disagreements between re-trainings of the same model due to factors such as randomization in the parameter initialization and mini-batches -- even when the trained models all attain similar accuracies. Such prediction churn can be very undesirable in practice. In this paper, we present several baselines for reducing churn and show that training on soft labels obtained by adaptively smoothing each example's label based on the example's neighboring labels often outperforms the baselines on churn while improving accuracy on a variety of benchmark classification tasks and model architectures.

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。

↑