List your company
ResearchFormerly WizWand

Untargeted white-box adversarial attack on FMNIST original (test)

0.1Original Error RateOriginal network
0.097080.116790.13650.15621Apr 12, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Original networkNetwork structure=CONV, Attack epsilon (ε)=0.042022.040.10.197
Original networkNetwork structure=CONV, Attack epsilon (ε)=0.082022.040.10.413
Original networkNetwork structure=CONV, Attack epsilon (ε)=0.062022.040.1030.283
Original networkNetwork structure=CONV, Attack epsilon (ε)=0.12022.040.1030.523
Original networkNetwork structure=FC, Attack epsilon (ε)=0.12022.040.1170.413
Original networkNetwork structure=FC, Attack epsilon (ε)=0.042022.040.130.207
Original networkNetwork structure=FC, Attack epsilon (ε)=0.072022.040.130.297
Robust networkNetwork structure=FC, Attack epsilon (ε)=0.07, Robust training epsilon (ε)=0.012022.040.1330.24
Original networkNetwork structure=FC, Attack epsilon (ε)=0.1142022.040.1330.497
Robust networkNetwork structure=FC, Attack epsilon (ε)=0.04, Robust training epsilon (ε)=0.012022.040.1360.2
Robust networkNetwork structure=FC, Attack epsilon (ε)=0.114, Robust training epsilon (ε)=0.012022.040.1360.387
Robust networkNetwork structure=CONV, Attack epsilon (ε)=0.04, Robust training epsilon (ε)=0.0052022.040.1370.22
Robust networkNetwork structure=FC, Attack epsilon (ε)=0.1, Robust training epsilon (ε)=0.012022.040.140.32
Robust networkNetwork structure=CONV, Attack epsilon (ε)=0.1, Robust training epsilon (ε)=0.0052022.040.1670.343
Robust networkNetwork structure=CONV, Attack epsilon (ε)=0.08, Robust training epsilon (ε)=0.0052022.040.170.257
Robust networkNetwork structure=CONV, Attack epsilon (ε)=0.06, Robust training epsilon (ε)=0.0052022.040.1730.23