Text Classification on AG-News (Adversarial Robustness Metrics)
85.7Logit FlippingDeepSet-TM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DeepSet-TMHeterogeneity (α)=0.5, Clients (n)=17, Adversaries (f)=42025.09 | 85.7 | 81.6 | 79.2 | 77.5 | 80.1 | 83.2 | 77.5 | |
| MeanHeterogeneity (α)=0.5, Clients (n)=17, Adversaries (f)=42025.09 | 84.5 | 81.4 | 81.2 | 76.4 | 72.6 | 54.9 | 54.9 | |
| CWTMHeterogeneity (α)=0.5, Clients (n)=17, Adversaries (f)=42025.09 | 84.3 | 81.4 | 80.2 | 76.4 | 72.6 | 55.3 | 55.3 | |
| GMHeterogeneity (α)=0.5, Clients (n)=17, Adversaries (f)=42025.09 | 83 | 80.7 | 80.9 | 76.6 | 74 | 52.6 | 52.6 | |
| CWMedHeterogeneity (α)=0.5, Clients (n)=17, Adversaries (f)=42025.09 | 78.9 | 78.4 | 75.9 | 74.1 | 74.4 | 53.2 | 53.2 |