SPHERE is a Python toolkit under development for forecasting of epidemiological dynamics in real-time. This project focuses on integrating practical Bayesian filtering methods and versatile forecasting tools, and implemented using Jax. Here are some key features we would like to have.
Key Features
Bayesian Filters: We plan to include various Bayesian filters like Particle Filter, Particle Markov Chain Monte Carlo (PMCMC), Ensemble Kalman Filter (EnKF), Extended Kalman Filter (EKF), and gradient-based filters.
Dynamical System Simulation: The package will be designed to simulate different dynamical systems, accommodating a range of non-Gaussian and non-linear properties, and generating synthetic data with varying complexities to support model testing and validation.
Forecasting Capabilities: We aim to integrate spatial-temporal regression analysis, trend detection methods, and deep learning frameworks such as Transformers to enhance forecasting accuracy.
This project is currently under active development. The features listed are part of our planned roadmap, and we are working towards implementing them in future releases.
When complete, the package will be suitable for:
Research: Providing tools for in-depth epidemiological research and analysis.
Epidemic Monitoring: Assisting in the real-time tracking and prediction of epidemic spread.
As the project is under construction, installation instructions and usage documentation will be provided in future updates.
We will provide comprehensive documentation with tutorials and examples once the initial release is ready. Please check back later for updates.
This project will be licensed under the MIT License - see the LICENSE file for more details once the project reaches a stable release.