CDCD (CNN-based Data-Model Co-Design) is the method in the work, "CNN-based Data-Model Co-Design for Efficient Test-termination Prediction", proposed in 2022 IEEE European Test Symposium (ETS).

In this example, the chip name is fixed as "x1" and "pair".
We have generated datasets named "x1" and "pair".

Generate datasets:

python my_utils.py -c x1 pair (--type ssl msl and or dom fe)
Generate ./pic folder and generate pictures

python getlabel_ma.py -c x1 pair (--type ssl msl and or dom fe)
Generate lables_ma.txt in './pic/chip(X)/fault(Y)/(Z)_resp/'

python modify_label.py -c x1 pair (--type ssl msl and or dom fe) (-t 0.899999)
Modify the labels into 0/1

python buildDataset.py -c x1 pair (--type ssl msl and or dom fe)
Build the dataset from the pic folder and divide it by 9:1 (trainset : testset)

-c: chip name
--type: fault type (default:["and", "or", "fe", "dom", "ssl", "msl"])
-t: threshold for modifying ma labels

---

Decision Tree:

python DAC4.py -s x1 -t pair
Decision Tree, trained with chip x1's trainset, tested with chip pair's testset

-s: source data
-t: target data

CDCD with DANN:

python main.py -s x1 -t pair
DANN, trained with chip x1's trainset with labels and chip pair's trainsset without labels, tested with chip pair's testset

-s: source data
-t: target data