  optimizing blood ...
    q=2 K=4  train_gap -0.027->-0.009  test_gap -0.027->-0.020  PQK-RBF +0.004->+0.007  bestAcc 0.774->0.765
  optimizing diabetes ...
    q=3 K=6  train_gap -0.100->-0.001  test_gap -0.068->-0.003  PQK-RBF -0.086->-0.046  bestAcc 0.737->0.697
  optimizing ilpd ...
    q=2 K=6  train_gap -0.008->-0.002  test_gap -0.004->-0.003  PQK-RBF +0.003->-0.005  bestAcc 0.715->0.708
  optimizing breast_cancer ...
    q=2 K=6  train_gap -0.458->+0.000  test_gap -0.424->-0.001  PQK-RBF -0.114->+0.000  bestAcc 0.942->0.736
  optimizing ionosphere ...
    q=4 K=6  train_gap -0.027->+0.100  test_gap -0.049->+0.031  PQK-RBF -0.077->-0.034  bestAcc 0.915->0.906
  optimizing sonar ...
    q=3 K=6  train_gap -0.154->+0.002  test_gap -0.066->+0.003  PQK-RBF -0.087->+0.005  bestAcc 0.712->0.625

Saved -> results/fig_feature_opt.png , results/study_feature_opt.csv
