This github repository is for the paper at ETS'24 - Fault Sensitivity Analysis of Printed Bespoke Multilayer Perceptron Classifiers
cite as
Fault Sensitivity Analysis of Printed Bespoke Multilayer Perceptron Classifiers
Pal, P.; Afentaki, F.; Zhao, H.; Saglam, G.; Hefenbrock, M.; Zervakis, G.; Beigl, M.; Tahoori, M. B.
2024 IEEE European Test Symposium (ETS), IEEE, 2024
Usage of the code:
The code can be simply run by command line through:
$ python3 experiment.py --DATASET 00 --SEED 00 --e_train 0.1 --dropout 0.1 --projectname FaultAnalysis
where the index of DATASET ranges from 00 to 12, SEED refers to random seed which was 00 - 09 in the experimental setup, e_train is the variation in the variation-aware training which was {0.0, 0.05, 0.1} in the experiment, while dropout was choosen from {0.0, 0.05, 0.1} in the experiment as well. The projectname would be the name of the folder that stores generated files during training, it can be modified as wanted.
The code for evaluation and visualiztion can be found in e.g., ./FaultAnalysis/Visualization.ipynb and ./FaultAnalysis/evaluation.ipynb