This covers the in-domain Bayesian deep learning experiments in Section 4.2.1.
- SGD (MAP):
bash train_sgd.sh(5 runs, seed 0-4) - BBB:
bash train_bbb.sh(5 runs, seed 0-4) - MC dropout:
bash train_mcdrop.sh(5 runs, seed 0-4) - SWAG:
bash train_swag.sh(5 runs, seed 0-4) - IVON:
bash train_ivon.sh(5 runs, seed 0-4) - Deep ensemble:
bash train_for_deep_ensemble.sh(runs SGD for seeds 5-24) - Multi-IVON:
bash train_for_multi_ivon.sh(runs IVON for seeds 5-24)
- SGD (MAP):
bash test_sgd.sh - BBB:
bash test_bbb.sh - MC dropout:
bash test_mcdrop.sh - SWAG:
bash test_swag.sh - IVON:
bash test_ivon.sh - Deep ensemble:
bash test_deepensemble.sh - Multi-IVON:
bash test_multiivon.sh