Code implementations for "Adaptive Uncertainty Estimation via High-Dimensional Testing on Latent Representations"
Trainers training our method and some of the baselines are provided in ./trainers. Other baselines are included in ./baselines. Users can customize their own trainers and import them in main.py.
Both the BNN and frequentist encoder architectures (e.g., LeNet) can be found in the ./models directory. Users can also customize own encoder architectures.
Create a yaml config file in the ./configs directory (examples can be found in the same directory), and run the following codes to run an experiment
python main.py
Results will be saved in ./checkpoints.