OpenTS-DL: Deep Learning Models
OpenTS-DL includes deep learning models for different tasks, including forecasting, probabilistic forecasting, and anomaly detection, while addressing different challenges in AGREE.
Robustness
K2VAE: A Koopman-Kalman Enhanced Variational AutoEncoder for Probabilistic Time Series Forecasting
International Conference on Machine Learning (ICML), 2025.
CATCH: Channel-Aware Multivariate Time Series Anomaly Detection via Frequency Patching
International Conference on Learning Representations (ICLR), 2025.
DUET: Dual Clustering Enhanced Multivariate Time Series Forecasting
ACM Knowledge Discovery and Data Mining (SIGKDD), 2025.
Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting
International Conference on Learning Representations (ICLR), 2024.
MM-Path: Multi-modal, Multi-granularity Path Representation Learning
ACM Knowledge Discovery and Data Mining (SIGKDD), 2025.
Weakly Guided Adaptation for Robust Time Series Forecasting
International Conference on Very Large Databases (PVLDB), 2023.
Explainability
Air-DualODE: Air Quality Prediction with Physics-guided Dual Neural ODEs in Open Systems.
International Conference on Learning Representations (ICLR), 2025.
Efficiency
Enhancing Diversity for Data-free Quantization
ACM Knowledge Discovery and Data Mining (SIGKDD), 2025.
Towards Lightweight Time Series Forecasting: a Patch-wise Transformer with Weak Data Enriching
International Conference on Data Engineering (ICDE), 2025.
Others
Position: What Can Large Language Models Tell Us about Time Series Analysis
International Conference on Machine Learning (ICML), 2024.
A Memory Guided Transformer for Time Series Forecasting
International Conference on Very Large Databases (PVLDB), 2024.
Multiple Time Series Forecasting with Dynamic Graph Modeling
International Conference on Very Large Databases (PVLDB), 2024.