我们采用 SoftCLT 对 TS2Vec(Yue 等,2022)进行单变量 TS 异常检测(AD)任务的实验,实验有两种不同的设置:正常设置是将每个数据集按照时间顺序分成两半,分别用于训练和评估;冷启动设置是在 UCR 档案中的 FordA 数据集上预训练模型,然后在每个数据集上进行评估。作为基线方法,我们考虑正常设置下的 SPOT(Siffer 等...
Self-Time(Fan 等人,2020 年)通过将同一 TS 的增强样本定义为正样本和负样本来捕捉 TS 之间的样本间关系,并通过解决分类任务来捕捉 TS 内部的时间关系,其中类标签使用子序列之间的时间距离来定义。TNC(Tonekaboni 等人,2021 年)使用正态分布定义窗口的时间邻域,并将邻域内的样本视为正样本。TS-SD(Shi 等人,202...
soft contrastive learning, specifically designed to capture the multifaceted nature of state behaviors more accurately. The data augmentation process enriches the dataset with varied representations of normal states, while soft contrastive learning fine-tunes the model's sensitivity to the subtle differences...
prototype deep-learning clustering transformer segmentation metric-learning fcn semantic-segmentation nonparametric softmax nearest-neighbours-classifier Updated Jun 30, 2022 Python Stonesjtu / Pytorch-NCE Star 318 Code Issues Pull requests The Noise Contrastive Estimation for softmax output written in ...
Recent unsupervised domain adaptation methods in medical image segmentation adopt centroid/prototypical contrastive learning (CL) to match the source and target features for their excellent ability of representation learning and semantic feature alignmen
Deep contextual reinforcement learning algorithm for scalable power scheduling Awol Seid Ebrie, Chunhyun Paik, Yongjoo Chung, Young Jin Kim Article 112243 Article preview select article An implicit aspect-based sentiment analysis method using supervised contrastive learning and knowledge embedding Research ...
Large-scale Distributed Evolutionary Reinforcement Learning (The distributed training module is suspended for maintenance) ATC,BYOL * denotes the features that we implemented. Supported Environments Gym, PyBullet and Unity environments with ML-Agents. ...
Contrastive loss Chen2014Deep and triplet lossF2015FaceNet need to carefully select instance pairs and triplet instances as the input of network in the train stage, since the performance of CNNs heavily depends on selected training instances. Similar to contrastive loss and triplet loss in ...
A novel model called deep contrastive mutual learning (DCML) for COVID-19 recognition was proposed in [19]. In this work, to avoid the overfitting, the training set was enriched using a multi-way data augmentation strategy based on the Fast AutoAugment algorithm [20]. Then, the contrastive ...
Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator. In Proceedings of the International Conference on Learning Representations Deep Inverse Workshop (ICLR-W), Virtual, 6–12 December 2020. [Google Scholar] Bao, H.; Dong, L.; Wei, F. ...