##Semi-supervised Approach for Dominant Gene Selection and Classification

SADGSC: Running the Code

The main script (main_SADGSC.m) runs the Semi-Supervised Dominant Gene Selection and Classification (SADGSC) algorithm.

Key Parameters and Variables

• cv_num : Number of cross-validation folds (default: 5)
• lambda1–lambda6 : Regularization and weighting parameters controlling sparsity, noise suppression, reconstruction error, and other model effects
• percentages : Proportion of unlabeled data used in semi-supervised training (e.g., [0.9, 0.7, 0.5]) (0.9 means 10% labeled)
• dataset_files : List of dataset .mat files to process

Workflow

1. Load datasets and prepare feature and target matrices.
2. Normalize the data and add bias term.
3. For each dataset and labeled data percentage:
   • Generate cross-validation splits (generateCVSet).
   • Mask a portion of the labels for semi-supervised learning (mask_target_entries).
   • Train the model using [out, ~] = model(...).
   • Predict outputs and assign labels using assignLabelsToHighestValue.
   • Evaluate performance: Average Precision, Weighted F1 Score, AUC.
4. Average results across CV folds and save to Excel (WeightedF1Scores.xlsx).

Example Usage

cv_num = 5;
percentages = [0.9, 0.7, 0.5];
dataset_files = {'breast_can.mat'};
lambda1 = 1; lambda2 = 1; lambda3 = 100;
lambda4 = 10; lambda5 = 1; lambda6 = 0.01;

% Run main script
run('main_SADGSC.m')