The laplace package facilitates the application of Laplace approximations for entire neural networks, subnetworks of neural networks, or just their last layer.
The package enables posterior approximations, marginal-likelihood estimation, and various posterior predictive computations.
It includes both weight-space and function-space Laplace approximations.
There is also a corresponding paper, Laplace Redux — Effortless Bayesian Deep Learning, which introduces the library, provides an introduction to the Laplace approximation, reviews its use in deep learning, and empirically demonstrates its versatility and competitiveness. Please consider referring to the paper when using our library:
@inproceedings{laplace2021,
title={Laplace Redux--Effortless {B}ayesian Deep Learning},
author={Erik Daxberger and Agustinus Kristiadi and Alexander Immer
and Runa Eschenhagen and Matthias Bauer and Philipp Hennig},
booktitle={{N}eur{IPS}},
year={2021}
}The code to reproduce the experiments in the paper is also publicly available; it provides examples of how to use our library for predictive uncertainty quantification, model selection, and continual learning.
Important
As a user, one should not expect Laplace to work automatically. That is, one should experiment with different Laplace's options (hessian_factorization, prior precision tuning method, predictive method, backend, etc!). Try looking at various papers that use Laplace for references on how to set all those options depending on the applications/problems at hand.
To install laplace with pip, run the following:
pip install laplace-torchTo use changes in this repository before they appear in a published release, clone the repository and install it from the checkout:
pip install -e .Additionally, if you want to use the asdfghjkl backend, please install it via:
pip install git+https://[email protected]/wiseodd/asdl@asdfghjklTip
Check out https://aleximmer.github.io/Laplace for more usage examples and API reference.
In the following example, a pre-trained model is loaded,
then the Laplace approximation is fit to the training data
(using a diagonal Hessian approximation over all parameters),
and the prior precision is optimized with cross-validation "gridsearch".
After that, the resulting LA is used for prediction with
the "probit" predictive for classification.
from laplace import Laplace
# Pre-trained model
model = load_map_model()
# User-specified LA flavor
la = Laplace(model, "classification",
subset_of_weights="all",
hessian_structure="diag")
la.fit(train_loader)
la.optimize_prior_precision(
method="gridsearch",
pred_type="glm",
link_approx="probit",
val_loader=val_loader
)
# User-specified predictive approx.
pred = la(x, pred_type="glm", link_approx="probit")Pull requests are very welcome. Please follow the guidelines in https://aleximmer.github.io/Laplace/devs_guide
Warning
AI-assisted contributions are welcome, but you must understand the problem you are solving. You must also understand the solutions your AI agent proposes and review them yourself. When opening a PR, you must write the description yourself. First-time contributors must disclose their AI usage and confirm that they understand the changes.
