Ridge Regression on Spiked covariance model N=500 synthetic
1.15Median Out-of-Sample MSEoptimization-based λmin
Evaluation results
| Method | Links | |
|---|---|---|
| optimization-based λminX-setting=fixed, aspect ratio (d/N)=0.9, noise (epsilon)=12026.05 | 1.15 | |
| model-based λfp (Alg. 2)X-setting=fixed, aspect ratio (d/N)=0.9, noise (epsilon)=12026.05 | 1.15 | |
| sample-based λsfp (Alg. 3)X-setting=fixed, aspect ratio (d/N)=0.9, noise (epsilon)=12026.05 | 1.19 | |
| signal-to-noise λSNX-setting=fixed, aspect ratio (d/N)=0.9, noise (epsilon)=12026.05 | 1.37 | |
| default λ0 = 1X-setting=fixed, aspect ratio (d/N)=0.9, noise (epsilon)=12026.05 | 1.61 |