Skip to content

Code source of "On Margins and Generalisation for Voting Classifiers" NeurIPS 2022 paper https://arxiv.org/abs/2206.04607

Notifications You must be signed in to change notification settings

vzantedeschi/dirichlet-margin-bound

Repository files navigation

Dependencies

Install PyTorch, following the guidelines. Then install the requirements:

pip3 install -r requirements.txt

Optimize PAC-Bayesian Bounds

python3 real_PACB.py dataset=MNIST num_trials=1 training.gamma=0.01

Default configuration are stored in 'config/real.yaml'.

Evaluate margin Bounds

python3 real_margin.py dataset=MNIST num_trials=1 training.gamma=0.01

Default configuration are stored in 'config/grid_search.yaml'.

To reproduce the results of the paper run

bash PACB.sh
bash margin.sh

About

Code source of "On Margins and Generalisation for Voting Classifiers" NeurIPS 2022 paper https://arxiv.org/abs/2206.04607

Resources

Stars

Watchers

Forks

Packages

No packages published