Martingale Posterior Neural Networks for Fast Sequential Decision Making

Citation

@article{duran2025scalable,
  title={Martingale Posterior Neural Networks for Fast Sequential Decision Making},
  author={Duran-Martin, Gerardo and S{\'a}nchez-Betancourt, Leandro and Cartea, {\'A}lvaro and Murphy, Kevin},
  journal={Advances in Neural Information Processing Systems},
  volume={38},
  year={2025}
}

Overview

The code is divided into notebooks and scripts. To reproduce the experiments, run the following notebooks

Reproducing Bayesian optimisation results

Run bayesopt-classic notebook.

Reproducing recommender system results

Run recommender-systems notebook.

Reproducing MNIST for classification results

Run mnist-online-classification notebook.

Reproducing MNIST as a neural contextual bandit problem

Run mnist-bandit-analysis notebook.

Run in-between uncertainty results

Run toy-in-between notebook.

Dependencies

bayesian-optimization==2.0.3
click==8.1.8
distrax==0.1.5
flax==0.10.3
jax==0.5.1
jaxlib==0.5.1
toml==0.10.2
numpy==2.2.3
vbll==0.4.6
matplotlib==3.10.0