Federated learning experiments

Setup

This README assumes that TensorFlow Federated and all other requirements are appropriately configured. See the instructions on the TFF website for more details.

LARC dataset

This repo uses the University of Michigan Learning Analytics Architecture sample dataset.

From the raw sample dataset tables, construct a single joined example dataset:

export LARC_MOCK="data/larc-mock/larc-mock.csv"

python scripts/generate_combined_larc_dataset.py \
    --prsn_identifying_info_fp data/larc-mock/PRSN_IDNTFYNG_INFO.csv \
    --stdnt_info_fp data/larc-mock/STDNT_INFO.csv \
    --stdnt_term_class_info_fp data/larc-mock/STDNT_TERM_CLASS_INFO.csv \
    --stdnt_term_info_fp data/larc-mock/STDNT_TERM_INFO.csv \
    --stdnt_term_trnsfr_info_fp data/larc-mock/STDNT_TERM_TRNSFR_INFO.csv \
    --out_fp $LARC_MOCK

This generates a single complete dataset called larc-mock.csv by joining the tables.

Next, we can train an example federated model from this dataset:

python train_feded_model.py --data_fp $LARC_MOCK