Code for the paper "Optimal Convergence for Agnostic Kernel Learning with Random Features"

Environments

• Matlab 2023a

Folders

• ./data contains processed datasets in .mat files.
• ./results records the variables in experiments and final results printed in .pdf files.

How to RUN

• Tuning the hyperparameter $\sigma^2$: Run run_select_sigma.m.
• Tuning the hyperparameter $\lambda$: Run run_select_lambda.m.
• Comparison Experiment
  • Run run_rf_leverage.m to obtain figures in Figs. 4 - 5.
  • Run draw_figs.m to illustrate results w.r.t. $M$ in Fig. 6.
  • Run run_table.m to obtain results in TABLE II.