Risk Minimization on Simulated Linear Regression N=10000, p=1000
-0.633Empirical RiskRSPG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RSPGlambda=10^-22018.02 | -0.633 | -0.632 | 75.1 | |
| RSPGlambda=10−3, Time=75.22018.02 | -0.629 | -0.628 | — | |
| SGDlambda=10^-2, mini-batch size=12018.02 | -0.322 | -0.322 | 96.1 | |
| SGDMini-batch size=1, lambda=10−3, Time=95.92018.02 | -0.162 | -0.155 | — | |
| SGDMini-batch size=10, lambda=10−3, Time=73.22018.02 | 1.1 | 1.45 | — | |
| SGDlambda=10^-2, mini-batch size=102018.02 | 1.36 | 1.37 | 73.4 | |
| SGDlambda=10^-2, mini-batch size=302018.02 | 2.61 | 2.61 | 71.6 | |
| SGDlambda=10^-2, mini-batch size=502018.02 | 3.08 | 3.08 | 409 | |
| SGDMini-batch size=30, lambda=10−3, Time=71.42018.02 | 4.79 | 5.02 | — | |
| SGDMini-batch size=50, lambda=10−3, Time=71.12018.02 | 6.03 | 6.21 | — |