vis_ts_classical_gatn_blackbox.py

import numpy as np
import tensorflow as tf

from models.timeseries import gatn, classical, classical_distilled
from utils.ts_classical.gatn_utils import visualise_gatn, visualise_gatn_dtw, train_scores_gatn, test_scores_gatn

tf.enable_eager_execution()
np.random.seed(0)


if __name__ == '__main__':
    # Select the model function which builds the model
    # Ensure that either :
    # 1) You set the model name in the constructor of the Model
    # 2) You set the model name in the train_base function
    atn_model_fn = gatn.TSFullyConnectedGATN

    # Select the model function which builds the model
    # Ensure that either :
    # 1) You set the model name in the constructor of the Model
    # 2) You set the model name in the train_base function
    clf_model_fn = classical.DTW

    # Select the model function which builds the model
    # Ensure that either :
    # 1) You set the model name in the constructor of the Model
    # 2) You set the model name in the train_base function
    student_model_fn = classical_distilled.TSDistilledLeNet5

    # train the model on a dataset
    dataset = 0

    # Which class to select as target
    TARGET_CLASS = 0

    CLASS_ID = None
    SAMPLE_ID = 0

    DATASET_TYPE = 'train'

    # Model name (used iff not None). Defaults to Model.name if this is None.
    clf_model_name = 'gridsearch-blackbox-' + clf_model_fn.__name__
    atn_model_name = 'gridsearch-blackbox-' + atn_model_fn.__name__
    student_model_name = 'gridsearch-blackbox-' + student_model_fn.__name__

    dataset = 'ucr/%s' % (str(dataset))

    if DATASET_TYPE == 'train':
        mse, acc_white, acc_black, rate, ids = train_scores_gatn(atn_model_fn, clf_model_fn, student_model_fn,
                                                                 dataset, TARGET_CLASS,
                                                                 atn_name=atn_model_name,
                                                                 clf_name=clf_model_name,
                                                                 student_name=student_model_name,
                                                                 shuffle=False)

    else:
        mse, acc_white, acc_black, rate, ids = test_scores_gatn(atn_model_fn, clf_model_fn, student_model_fn,
                                                                dataset, TARGET_CLASS,
                                                                atn_name=atn_model_name,
                                                                clf_name=clf_model_name,
                                                                student_name=student_model_name,
                                                                shuffle=False)

    print("Adversarial IDS : ", ids)

    visualise_gatn(atn_model_fn, clf_model_fn, student_model_fn, dataset, TARGET_CLASS, class_id=CLASS_ID, sample_id=SAMPLE_ID,
                   plot_delta=True, atn_name=atn_model_name, clf_name=clf_model_name, student_name=student_model_name,
                   dataset_type=DATASET_TYPE, save_image=True)

    visualise_gatn_dtw(atn_model_fn, clf_model_fn, student_model_fn, dataset, TARGET_CLASS, class_id=CLASS_ID, sample_id=SAMPLE_ID,
                       plot_delta=False, atn_name=atn_model_name, clf_name=clf_model_name, student_name=student_model_name,
                       dataset_type=DATASET_TYPE, save_image=True)