====== Assignment: Nested Cross-Validation ======

**📅 Deadline:** 22.5.2026 21:59

**🏦 Points:** 4
===== Task Description =====
In this assignment, you will implement a nested cross-validation procedure and use it to train an RBF-kernel SVM classifier. You will also estimate its generalization performance by constructing a confidence interval. The full assignment details are available in the {{ :courses:becm33mlf:homeworks:ls26_hw_nested_cv.pdf |Assignment PDF}}.

You are provided with a {{ :courses:becm33mlf:homeworks:ls26_hw_nestedcv_template.zip | template }} containing the following files:

  * **train_svm.py**: This file includes the functions **main**, **select_best_c** that you are required to implement.
  * **main.py**: This file evaluates performance of the SVM classifier and validatity of the confidence interval.
  * **utils.py**: Contains helper functions for loading and saving data, and visualizations. You do not need to modify this file.
  * **X_train.npy**: Contains development inputs.
  * **y_train.npy**: Contains development labels.
  * **test-cases**: A folder containing public test cases to help you verify your implementation before submitting to [[https://cw.felk.cvut.cz/brute/student/course/BECM33MLF/mle|BRUTE]].

Your objective is to implement the function **main**, which performs the nested CV, including C tuning, SVM training and CI computation, and **select_best_c.py** which selects the regularization parameter C via the cross-validation.

<note important>
All python files must be stored in the root of the .zip sent for submission.
</note>

===== How to Test =====
After completing your implementation, you can test your solution using the following commands before submitting it to BRUTE:

----


== Test Case ==
  python main.py test-cases/public/instances/public_test.json
Expected output:

  Loading model and scaler...
  Model, scaler, and results loaded successfully.
  X shape: (3804, 10)
  Scaling hidden features...
  Making predictions...
  Error on Hidden Test Set: 0.1338
  CV Error: 0.1269, Computed CI: [0.1133, 0.1404]
  Reference Error: 0.1338, Reference CI: [0.1133, 0.1404]
  Test OK
----    


===== Submission Guidelines =====
  * Submit the following files as a .zip via BRUTE:


  cv_results.json
  feature_scaler.joblib
  main.py
  rbf_svm_model.joblib
  utils.py


  * All files must be stored in the root of the .zip sent for submission.
  * Make sure your implementation passes the test cases provided above. Good luck! 😊