Robust Halfspace Learning on Log-concave mixtures Theoretical bounds
1Margin γTalwar
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TalwarNoise model=adversarial label, malicious, Marginal=log-concave mix., Cert. degree=22026.06 | 1 | 1 | — | |
| ShenNoise model=malicious, Marginal=log-concave mix., Cert. degree=22026.06 | 1 | 1 | — | |
| Zeng and ShenNoise model=malicious, sparse, Marginal=log-concave mix., Cert. degree=22026.06 | 1 | 1 | — | |
| degree-2t reweighted hingeNoise model=malicious, nasty, Marginal=log-concave mix., Cert. degree=2t (tunable)2026.06 | 1 | 0 | 2 |