df_compare_f1.py

def compare_f1(hyperparams, rounds=1):
    local_mean_f1_scores = []
    local_max_f1_scores = []
    fed_avg_f1_scores = []
    for i in range(0, rounds):
        fed_avg = FedAvg(**hyperparams)
        fed_avg.fit(X_train, y_train)
        preds = fed_avg.predict(X_test)
        fed_avg_f1_scores.append(f1_score(y_test, preds, average='weighted'))
        tmp = []
        for model in fed_avg.models:
            local_pred = model.predict(X_test)
            tmp.append(f1_score(y_test, local_pred, average='weighted'))
        local_mean_f1_scores.append(np.mean(tmp))
        local_max_f1_scores.append(np.max(tmp))
    return np.mean(local_mean_f1_scores), np.mean(local_max_f1_scores), np.mean(fed_avg_f1_scores)