Published January 14, 2026 | Version 1.0.0
Software Open

Clinically Constrained Reinforcement Learning for Optimized Adaptive Data Augmentation in Chest X-Ray Pneumonia Classification

Authors/Creators

  • 1. Department of Information Systems, Faculty of Computers and Artificial Intelligence, Matrouh University, Egypt

Description

A clinically constrained reinforcement learning framework for chest X-ray pneumonia classification. Combines CNN-Transformer architecture with PPO-based adaptive augmentation, achieving 97.23% accuracy with excellent calibration (ECE = 0.016).

Files

mmahmoudai/pneumonia-classification-ppo-transfor-cust-cnn-v1.zip

Files (40.4 kB)

Additional details