run_experiment.py

import time

import nest_asyncio

from lnb.data_prep import load_data, split_data
from lnb.distance import compute_achilles, top_n_vulnerable_records
from lnb.mia import mia

nest_asyncio.apply()


def run_exp():
    path_to_data = "data/adult/Adult_dataset.csv"
    path_to_metadata = "data/adult/Adult_metadata_discretized.json"

    df, categorical_cols, continuous_cols, meta_data = load_data(
        path_to_data, path_to_metadata
    )
    df_aux, df_eval, df_target = split_data(
        df, "data/adult/1000_indices.pickle"
    )

    print("Calculating Achilles scores...")
    t1 = time.time()
    all_dists = compute_achilles(
        df_target, categorical_cols, continuous_cols, meta_data, 5
    )
    t2 = time.time()

    print(f"time taken = {(t2-t1)}")

    top_n_records = top_n_vulnerable_records(all_dists, 100)

    t1 = time.time()

    mia(
        path_to_data=path_to_data,
        path_to_metadata=path_to_metadata,
        path_to_data_split="data/adult/1000_indices.pickle",
        target_records=top_n_records[0:1],
        generator_name="SYNTHPOP",
        n_original=1000,
        n_synth=1000,
        n_datasets=10,
        epsilon=0.0,
        output_path="./output/files/",
    )
    t2 = time.time()
    time_taken = t2 - t1

    print(f"time taken for MIAs = {time_taken}")


if __name__ == "__main__":
    run_exp()