具体来说,Transformers可以说是提取长序列中元素之间语义相关性的最成功的解决方案。然而,在时间序列建模中,我们要从连续点的有序集合中提取时间关系。在Transformers中采用位置编码和token嵌入子序列有助于保留一些序列信息,但排列不变(permutation-invariant)的自注意力机制的性质不可避免地导致了时间信息的丢失。 为了...
WaveForM: Graph Enhanced Wavelet Learning for Long Sequence Forecasting of Multivariate Time Series Temporal-Frequency Co-Training for Time Series Semi-Supervised Learning SEnsor Alignment for Multivariate Time-Series Unsupervised Domain Adaptation Time-aware Random Walk Diffusion to Improve Dynamic Graph Learn...
本次分享是今年AAAI 2023 顶会中时空数据挖掘相关的论文,目前共整理了23篇,有缺漏也欢迎大家评论区补充哈! 扫码添加小享,回复“时空数据” 免费获取全部论文+代码合集 1. GMDNet: A Graph-based Mixture Density Network for Estimating Packages' Multimodal Travel Time Distribution 标题:基于图的混合密度网络用于...
论文链接:https://www.microsoft.com/en-us/research/publication/learning-decomposed-spatial-relations-for-multi-variate-time-series-modeling/ 多变量时间序列建模(Multi-variate time-series modeling)在金融、气象、交通等领域都得到了广泛的应用。许多运用图与图神经网络来表征变量间空间关系的方法都取得了不俗的...
This work is accepted for publication in Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI 2023). MHCCL Overview: Abstract Learning semantic-rich representations from raw unlabeled time series data is critical for downstream tasks such as classification and forecasting. Contrastive...
AAAI invites proposals for the 2023 Spring Symposium Series, to be held at a TBD time/venue, in Palo Alto, California. The Spring Symposium Series is an annual set of meetings run in parallel at a common site. It is designed to bring colleagues together in an intimate forum while at the...
SimPSI: A Simple Strategy to Preserve Spectral Information in Time Series Data Augmentation(时序数据增强) 时空数据 Fully-Connected Spatial-Temporal Graph for Multivariate Time Series Data 多变量时间序列数据的全连接时空图 简述:论文提出了一种名为FC-STGNN的全连接时空图神经网络方法,用于多变量时间序列数据...
Detecting Multivariate Time Series Anomalies with Zero Known Label(AAAI 2023) This repository provides a PyTorch implementation of MTGFlow (Paper), which is the unsupervised anomaly detection and localization method. This repository is based on GANF. Framework Main results Requirements python==3.8.5 ...
论文标题:Enhancing Masked Time-Series Modeling via Dropping Patches 前言 读本文首先要了解掩码建模这种自监督预训练方法,既通过基于未掩码部分重建掩码内容来改进表征学习,促使模型学习到更具鲁棒性和通用性的表征,这种方法早先已广泛应用到CV和NLP等领域。
Causal Recurrent Variational Autoencoder for Medical Time Series Generation Improvement-Focused Causal Recourse (ICR)COCA: COllaborative CAusal Regularization for Audio-Visual Question Answering Direct Heterogeneous Causal Learning for Resource Allocation Problems in Marketing Self-supervised Learning ...