Simulations of physical systems are often slow and need lots of compute, which makes them unpractical for real-world applications like digital twins, or when they have to run thousands of times for sensitivity analyses. The goal of AutoEmulate is to make it easy to replace simulations with fast, accurate emulators. To do this, AutoEmulate automatically fits and compares various models, ranging from simple models like Radial Basis Functions and Second Order Polynomials to more complex models like Support Vector Machines, Gaussian Processes and Conditional Neural Processes to find the best emulator for a simulation.
The project is in early development.
There's currently a lot of development, so we recommend installing the most current version from GitHub:
pip install git+https://github.com/alan-turing-institute/autoemulate.gitThere's also a release available on PyPI (will not contain the most recent features and models)
pip install autoemulateFor contributors using Poetry:
git clone https://github.com/alan-turing-institute/autoemulate.git
cd autoemulate
poetry installimport numpy as np
from autoemulate.compare import AutoEmulate
from autoemulate.experimental_design import LatinHypercube
from autoemulate.simulations.projectile import simulate_projectile
# sample from a simulation
lhd = LatinHypercube([(-5., 1.), (0., 1000.)])
X = lhd.sample(100)
y = np.array([simulate_projectile(x) for x in X])
# compare emulator models
ae = AutoEmulate()
ae.setup(X, y)
best_model = ae.compare()
# training set cross-validation results
ae.print_results()
# plot cv results
ae.plot_results()
# test set results for the best model
ae.evaluate_model(best_model)
# refit on full data
best_emulator = ae.refit_model(best_model)
# emulate
best_emulator.predict(X)You can find tutorials, FAQs and the API reference here. The documentation is still work in progress.
 Kalle Westerling 📖 💻 🖋 |
 Bryan M. Li 💻 |
 martin 💻 🤔 📖 🚧 🔬 👀 |
 Eric Daub 🤔 📆 👀 💻 |
 steven niederer 🤔 🖋 📆 |
 Maximilian Balmus 💻 🐛 |
 Sophie Arana 🖋 📖 📆 |