Federated Unlearning on Fashion-MNIST
65.93Pre-training Accuracyf-FUM (KL-KL)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| f-FUM (KL-KL)N=10, Removed (%)=62026.02 | 65.93 | 77.1 | |
| f-FUM (KL-JS)N=10, Removed (%)=62026.02 | 65.93 | 79.9 | |
| f-FUM (KL-chi2)N=10, Removed (%)=62026.02 | 65.93 | 76.25 | |
| NoTN=10, Removed (%)=62026.02 | 65.93 | 26.67 | |
| HalimiN=10, Removed (%)=62026.02 | 65.93 | 77.82 | |
| f-FUM (KL-KL)N (Number of clients)=5, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 50.77 | 70.7 | |
| f-FUM (KL-JS)N (Number of clients)=5, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 50.77 | 77.85 | |
| f-FUM (KL-χ²)N (Number of clients)=5, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 50.77 | 67.8 | |
| NoTN (Number of clients)=5, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 50.77 | 68.4 | |
| HalimiN (Number of clients)=5, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 50.77 | 95.48 | |
| f-FUM (KL-KL)N (Number of clients)=10, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 42.35 | 67.67 | |
| f-FUM (KL-JS)N (Number of clients)=10, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 42.35 | 77.4 | |
| f-FUM (KL-χ²)N (Number of clients)=10, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 42.35 | 64.6 | |
| NoTN (Number of clients)=10, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 42.35 | 39.92 | |
| HalimiN (Number of clients)=10, Removed (%)=2%, Backbone=MINILENET, Unlearning Scenario=Confuse2026.02 | 42.35 | 68.52 | |
| f-FUM (KL-KL)N=10, Removed (%)=102026.02 | 41.98 | 49.2 | |
| f-FUM (KL-JS)N=10, Removed (%)=102026.02 | 41.98 | 70.48 | |
| f-FUM (KL-chi2)N=10, Removed (%)=102026.02 | 41.98 | 48.17 | |
| NoTN=10, Removed (%)=102026.02 | 41.98 | 32.13 | |
| HalimiN=10, Removed (%)=102026.02 | 41.98 | 79.58 |