Bayesian optimization interface for the laplace-torch library
Project description
Bayesian Optimization Interface for laplace-torch
Installation
Install PyTorch first, then:
pip install --upgrade laplace-bayesopt
Usage
Basic usage
from laplace_bayesopt.botorch import LaplaceBoTorch
def get_net():
# Return a *freshly-initialized* PyTorch model
return torch.nn.Sequential(
...
)
# Initial X, Y pairs, e.g. obtained via random search
train_X, train_Y = ..., ...
model = LaplaceBoTorch(get_net, train_X, train_Y)
# Use this model in your existing BoTorch loop, e.g. to replace BoTorch's SingleTaskGP model.
The full arguments of LaplaceBoTorch
can be found in the class documentation.
Check out examples in examples/
.
Useful References
- General Laplace approximation: https://arxiv.org/abs/2106.14806
- Laplace for Bayesian optimization: https://arxiv.org/abs/2304.08309
- Benchmark of neural-net-based Bayesian optimizers: https://arxiv.org/abs/2305.20028
- The case for neural networks for Bayesian optimization: https://arxiv.org/abs/2104.11667
Citation
@inproceedings{kristiadi2023promises,
title={Promises and Pitfalls of the Linearized {L}aplace in {B}ayesian Optimization},
author={Kristiadi, Agustinus and Immer, Alexander and Eschenhagen, Runa and Fortuin, Vincent},
booktitle={AABI},
year={2023}
}
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
laplace_bayesopt-0.1.6.tar.gz
(95.7 kB
view hashes)
Built Distribution
Close
Hashes for laplace_bayesopt-0.1.6-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 32c7baa5e71e62a369b47918c79f2347f20b4d4e46dfe534f865106bc2ad797d |
|
MD5 | 9ce6ec57bd2dc76b05f4010cb44a326e |
|
BLAKE2b-256 | ecff509288c88f4a3d39299d1ec096d57aa9b55e12b9d77dd3ddd2bcd114951f |