exploiting-ml-multimodality

Research of ML multimodality

How to run experiment end-to-end

Prepare data

python3 compas_arrest.py

Run experiment

Run python script "run_experiment.py". pass config name of sweep id to start training

python3 run_experiment.py --config test-config.yaml

Retreive models

1. Get models

python3 get_models_meta.py --config Lr.yaml

2. Filter models

python3 filter_models.py --config Lr.yaml --margin 0.01

3. Download models

python3 download_models.py --config Lr.yaml

Get predictions

python3 get_predictions.py --config test-config.yaml

python3 get_predictions_random.py --config Lr.yaml --sample 10

Combined importances

python3 get_models_meta.py --config Batches_churn.yaml &&
python3 filter_models.py --config Batches_churn.yaml --margin 0.01 &&
python3 download_models.py --config Batches_churn.yaml &&
python3 get_feature_importance.py --config Random_seed_churn.yaml

Combined predictions

python3 get_models_meta.py --config Batches.yaml &&
python3 filter_models.py --config Batches.yaml --margin 0.01 &&
python3 download_models.py --config Batches.yaml &&
python3 get_predictions.py --config Batches.yaml

Datasets:

1. ADULT dataset https://www.kaggle.com/datasets/wenruliu/adult-income-dataset Source http://www.cs.toronto.edu/~delve/data/adult/desc.html http://www.cs.toronto.edu/~delve/data/adult/adultDetail.html

2. Compas dataset (from paper github account)

3. Customer churn https://www.kaggle.com/datasets/barelydedicated/bank-customer-churn-modeling

4. Bank Marketing Dataset (Portugal) http://archive.ics.uci.edu/dataset/222/bank+marketing

5. Employee Churn (HR Analytics by IBM) https://www.kaggle.com/datasets/giripujar/hr-analytics

6. cancer+risk+factors https://archive.ics.uci.edu/dataset/383/cervical+cancer+risk+factors