Re-implementing MIMO

Final project for DD2412 Deep Learning, Advanced Course at KTH Royal Institute of Technology, Stockholm.

Group members

• Diogo Pinheiro (https://github.com/DiogorPinheiro)
• Jakob Lindén (https://github.com/jakobGTO)

Dependencies

1. Tensorflow
2. Numpy
3. Robustness-metrics (source: https://github.com/google-research/robustness_metrics)
4. Matplotlib
5. Uncertainty Baselines (source: https://github.com/google/uncertainty-baselines)

Running Uncertainty-Baseline comparison tests

1. Clone repository from https://github.com/google/uncertainty-baselines.git

2. Place script "run-tests" in the root directory of the cloned repository

3. Run command "chmod u+x run-tests" in that same directory

4. Run tests in root directory using

   ./run-tests --dataset cifar100 --model batchensemble --gpu True --download True --cores 4 --requirements False --epochs 10
   

   The following parameters are currently provided: - dataset : "cifar100" or "cifar10"

    - model : "all" (will run all pre-defined models) or write a specific one (e.g. "batchensemble" or "dropout")
    - gpu : Boolean that represents whether the process will be ran using gpu or not
    - download: If set to True it will download the required dataset to a temporary directory
    - cores : Number of cores to be user
    - requirements: If set to True it will first and foremost install all necessary dependencies
    - epochs : Number of epochs to be performed