thuml/Time-Series-Library
Time Series Library
This PyTorch-based deep learning library provides a framework for analyzing and forecasting temporal data. It implements specialized architectures for time series forecasting, anomaly detection, data imputation, and classification.
The project distinguishes itself through the inclusion of zero-shot inference capabilities, allowing large-scale temporal models to be evaluated on unseen datasets without requiring task-specific fine-tuning.
The framework covers a broad range of analytical capabilities, including the recovery of missing values in incomplete datasets, the identification of irregular patterns or outliers, and the prediction of future values across short and long term horizons. It also includes a modular model registry and unified dataset interfaces to support research reproducibility and performance benchmarking across standard datasets.
Features
- Time Series Forecasting - Provides a comprehensive framework of deep learning architectures for predicting future values in temporal data sequences.
- Deep Learning Forecasting - Applies deep neural networks to analyze historical trends for univariate and multivariate time series prediction.
- PyTorch-Based Frameworks - Provides a PyTorch-based deep learning framework specifically for temporal data analysis and forecasting.
- Sequential Pattern Analysis - Uses RNNs and CNNs to analyze and extract characteristic trends and patterns within sequential temporal data.
- Temporal Task Pipelines - Organizes distinct neural network pipelines tailored for forecasting, classification, imputation, and anomaly detection.
- Time Series Anomaly Detection - Identifies unusual patterns or outliers in sequential data that deviate from expected behavior using deep learning.
- Zero-Shot Forecasting - Generates time series predictions using pre-trained models without requiring task-specific training data.
- Zero-Shot Temporal Models - Implements a framework for evaluating large-scale temporal models on unseen datasets without task-specific fine-tuning.
- Zero-Shot Inference - Enables predictions on unseen datasets using pre-trained large models without task-specific parameter updates.
- Time Series Imputation - Fills gaps in incomplete temporal datasets using deep learning to recover missing values.
- Model Performance Benchmarks - Ranks multiple candidate model architectures based on standardized performance metrics across datasets.
- Custom Predictive Model Development - Provides a framework for developers to add new model architectures and define integration experiment scripts.
- Model Architecture Registries - Implements a consistent registry for managing diverse deep learning architectures and experiment scripts.
- Model Performance Benchmarking - Enables benchmarking deep learning architectures against standard datasets using centralized metrics to ensure research reproducibility.
- Evaluation Metrics - Provides standardized scripts and methods for measuring the accuracy of sequence learning models across datasets.
- Time Series Classification - Includes specialized neural network pipelines for assigning categorical labels to time series segments based on temporal patterns.
- Unified Dataset I/O - Abstracts various time series data sources into a common format for consistent model input.
- General Time Series Analysis - Comprehensive library for general time series analysis and modeling.
- Time Series Analysis - Deep learning-based time series analysis and forecasting.
- Analysis Toolkits - Comprehensive library of deep learning models for sequences.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
