摘要
一位新闻记者-机器人与机器学习的新闻编辑-每日新闻-关于人工智能的研究结果在一份新的报告中讨论。根据NewsRx编辑的《来自中国北京的新闻》报道，这项研究指出：“时间序列是许多领域的典型数据类型；然而，对大量时间序列数据进行标记可能是昂贵和耗时的。从未标记的时间序列数据中学习有效的表示是一项具有挑战性的任务。”新闻记者引用了北京邮电大学的一篇研究文章:“对比学习是获取无标记时间序列数据表征的一种很有前景的方法。因此,我们提出了一种基于时频融合的自监督时间序列表征学习框架,并与(TF-FC)进行了对比,从无标记时间序列数据中学习时间序列表征。”TF-FC将时域增强和频域增强相结合，产生不同的样本，对于时域增强，原始时间序列数据通过时域增强库(如抖动、缩放、置换、掩蔽)得到时域增强数据，对于频域增强，首先将原始时间序列经快速傅立叶变换(FFT)分析转换为频域数据，然后，频率数据经过频域增强库(如低通滤波器、去频、加频、相位滤波)得到频域增强数据,时域增强数据与频域增强数据的融合方法是核PCA,它有利于提取高维空间的非线性特征,通过对时间序列的时域和频域进行CAPTU环合,得到时域增强数据和频域增强数据。该方法能够从数据中提取更多的信息特征，增强了模型区分不同时间序列的能力。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Research findings on artificial intell igence are discussed in a new report. According to news originating from Beijing , People's Republic of China, by NewsRx editors, the research stated, "Time seri es is a typical data type in numerous domains; however, labeling large amounts o f time series data can be costly and time-consuming. Learning effective represen tation from unlabeled time series data is a challenging task." The news journalists obtained a quote from the research from Beijing University of Posts and Telecommunications: "Contrastive learning stands out as a promising method to acquire representations of unlabeled time series data. Therefore, we propose a self-supervised time-series representation learning framework via Time -Frequency Fusion Contrasting (TF-FC) to learn time-series representation from u nlabeled data. Specifically, TF-FC combines time-domain augmentation with freque ncy-domain augmentation to generate the diverse samples. For time-domain augment ation, the raw time series data pass through the time-domain augmentation bank ( such as jitter, scaling, permutation, and masking) and get time-domain augmentat ion data. For frequency-domain augmentation, first, the raw time series undergoe s conversion into frequency domain data following Fast Fourier Transform (FFT) a nalysis. Then, the frequency data passes through the frequency-domain augmentati on bank (such as low pass filter, remove frequency, add frequency, and phase shi ft) and gets frequency-domain augmentation data. The fusion method of time-domai n augmentation data and frequency-domain augmentation data is kernel PCA, which is useful for extracting nonlinear features in high-dimensional spaces. By captu ring both the time and frequency domains of the time series, the proposed approa ch is able to extract more informative features from the data, enhancing the mod el's capacity to distinguish between different time series."
NETL
NSTL