dbr7 / deepmutationpp_cnn_mutation_mnist

Important Files and Folders (in src folder)

• mnist_conv_train.py: The original MNIST model training file.
• deepcrime_scripts: It contains original DeepCrime scripts such as utils.py or stats.py
• cnn_operator.py and generator.py: These files are the core of DeepMutation++. I only added few lines of code to set the random seed.
• run_op_dmpp.py: This is the main file to call generator.py. After generating DeepMutation++ mutatns, it checkes the accuracies of generated mutants to determine the killing of them. The definition of the killing here is from DeepCrime paper.
• compute_ki.py: This computes the killing infomation for DeepMutation++ mutants. get_prediction_array_mutant and save_killing_info functions do this job.
• 0_subsumption_study.ipynb: The jupyter notebook to analyse the redundancy between mutants. There are two experiments: 1) whether DeepCrime mutants subsume DeepMutation++ mutants, 2) whether DeepMutation++ mutants subsume DeepCrime mutants.