Table 6 GPBEAM-DE attacking BEAM-related models. GPBEAM-DE-N denote the GPBEAM-DE with N genes in each individual. GPBEAM-DE-5* indicates GPBEAM-DE-5 without perturbation overflow
From: Perturbing BEAMs: EEG adversarial attack to deep learning models for epilepsy diagnosing
Attack Method | GPBEAM-DE-5* | GPBEAM-DE-5 | GPBEAM-DE-3 | GPBEAM-DE-1 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
\(\epsilon\) | 0.1 | 0.3 | 0.5 | 0.1 | 0.3 | 0.5 | 0.1 | 0.3 | 0.5 | 0.1 | 0.3 | 0.5 |
Acc | 0.93 | 0.73 | 0.66 | 0.64 | 0.25 | 0.15 | 0.81 | 0.37 | 0.21 | 0.89 | 0.73 | 0.65 |
SR | 0.02 | 0.23 | 0.30 | 0.31 | 0.70 | 0.80 | 0.23 | 0.59 | 0.74 | 0.05 | 0.23 | 0.31 |
\({\mathrm{DL}}_{B}\) | 0.01 | 0.04 | 0.07 | 0.06 | 0.18 | 0.29 | 0.04 | 0.12 | 0.19 | 0.02 | 0.05 | 0.07 |