Risk Minimization on Simulated Linear Regression N=10000, p=2000
-0.647Empirical RiskRSPG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RSPGlambda=10^-22018.02 | -0.647 | -0.646 | 117 | |
| RSPGlambda=10−3, Time=1172018.02 | -0.646 | -0.646 | — | |
| SGDlambda=10^-2, mini-batch size=12018.02 | -0.131 | -0.13 | 144 | |
| SGDMini-batch size=1, lambda=10−3, Time=1452018.02 | 0.187 | 0.194 | — | |
| SGDMini-batch size=10, lambda=10−3, Time=99.22018.02 | 0.428 | 0.431 | — | |
| SGDMini-batch size=30, lambda=10−3, Time=95.72018.02 | 0.479 | 0.481 | — | |
| SGDMini-batch size=50, lambda=10−3, Time=1662018.02 | 0.496 | 0.499 | — | |
| SGDlambda=10^-2, mini-batch size=102018.02 | 3.23 | 3.23 | 99.1 | |
| SGDlambda=10^-2, mini-batch size=302018.02 | 5.63 | 5.63 | 95.6 | |
| SGDlambda=10^-2, mini-batch size=502018.02 | 6.52 | 6.53 | 503 |