FFAD:一种利用傅里叶变换和自动编码器评估生成时间序列数据的新型指标
FFAD: A Novel Metric for Assessing Generated Time Series Data Utilizing Fourier Transform and Auto-encoder
AI总结:
针对时间序列生成数据缺少类似 FID 的标准评估指标,本文提出结合傅里叶变换、自动编码器与 Fréchet 距离的 FFAD,并验证其区分类别样本的能力。
摘要:
基于深度学习的生成模型在生成逼真图像、视频和音频方面的成功,引发了一个关键考量:如何有效评估合成样本的质量。尽管 Fréchet Inception Distance(FID)是图像合成中评估生成模型的标准指标,但在时间序列数据方面却明显缺乏可与之相比的指标。这种评估能力上的缺口源于缺少一个在基准时间序列数据集上预训练、并被广泛接受的特征向量提取器。为应对这些与时间序列质量评估相关的挑战,尤其是在 Fréchet Distance(Fréchet 距离)语境下,本研究提出了一种利用傅里叶变换和自动编码器的新型解决方案,称为 Fréchet Fourier-transform Auto-encoder Distance(FFAD)。通过实验结果,我们展示了 FFAD 有效区分来自不同类别样本的潜力。这一新型指标成为评估生成时间序列数据的基础工具,为持续改进基于深度学习的生成模型领域中的评估方法作出贡献。
英文摘要:
The success of deep learning-based generative models in producing realistic images, videos, and audios has led to a crucial consideration: how to effectively assess the quality of synthetic samples. While the Fréchet Inception Distance (FID) serves as the standard metric for evaluating generative models in image synthesis, a comparable metric for time series data is notably absent. This gap in assessment capabilities stems from the absence of a widely accepted feature vector extractor pre-trained on benchmark time series datasets. In addressing these challenges related to assessing the quality of time series, particularly in the context of Fréchet Distance, this work proposes a novel solution leveraging the Fourier transform and Auto-encoder, termed the Fréchet Fourier-transform Auto-encoder Distance (FFAD). Through our experimental results, we showcase the potential of FFAD for effectively distinguishing samples from different classes. This novel metric emerges as a fundamental tool for the evaluation of generative time series data, contributing to the ongoing efforts of enhancing assessment methodologies in the realm of deep learning-based generative models.
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。