https://github.com/probml/bandits
Bayesian Bandits
https://github.com/probml/bandits
Last synced: about 1 year ago
JSON representation
Bayesian Bandits
- Host: GitHub
- URL: https://github.com/probml/bandits
- Owner: probml
- License: mit
- Created: 2021-11-30T16:50:35.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2023-09-03T08:37:57.000Z (almost 3 years ago)
- Last Synced: 2025-03-28T10:47:36.225Z (about 1 year ago)
- Language: Jupyter Notebook
- Size: 11.2 MB
- Stars: 67
- Watchers: 10
- Forks: 7
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Efficient Online Bayesian Inference for Neural Bandits
**🚨Breaking changes🚨**
See [aistats2022 release](https://github.com/probml/bandits/releases/tag/aistats2022) For a [JSL@ceeef0](https://github.com/probml/JSL/commit/ceeef0f02b185c7188afb40b977a9406d97c21ba) and `jax<=0.2.22` compatible version.
----
By [Gerardo Durán-Martín](http://github.com/gerdm), [Aleyna Kara](https://github.com/karalleyna), and [Kevin Murphy](https://github.com/murphyk)
[Arxiv paper](https://arxiv.org/abs/2112.00195).
[Slides](https://probml.github.io/bandits/1)
-----
## Installation
```
pip install fire
pip install ml-collections
```
## Reproduce the results
There are two ways to reproduce the results from the paper
### Run the scripts
To reproduce the results, `cd` into the project folder and run
```bash
python bandits test
```
```bash
python bandits run_and_plot
```
### Step by step
If you only want to reproduce the results, run
```bash
python bandits run_experiments
```
If you have previously reproduced the results and want to reproduce the plots, run
```bash
python bandits plot_experiments
```
The results will be stored inside `bandits/figures/`.
### Execute the notebooks
An alternative way to reproduce the results is to simply open and run [`subspace_bandits.ipynb`](https://github.com/probml/bandits/blob/main/bandits/scripts/subspace_bandits.ipynb)