iamshibly/FedRL-FuseNet — reverse-engineered prompt

Build me a notebook first research project for FedRL FuseNet, a privacy aware and explainable system for classifying brain tumor MRI scans into four classes. I want the main workflow to cover contrast enhanced preprocessing, learning from raw, enhanced, and residual image information, an adaptive fusion step, reinforcement guided preset selection, federated training, evaluation, reliability analysis, and model inspection for explainability.

Please structure it like a reproducible research repo, with one main notebook and separate notebooks or readable reports for preprocessing validation, fusion ablations, federated learning ablations, RL UCB analysis, baseline and backbone comparisons, and external validation. Include GitHub viewable outputs where possible, plus a supplementary markdown document that summarizes the extra experimental results. If the original data or full explainability notebook is not included here, leave clear placeholders and instructions so someone can plug those in later. Keep it practical and easy to run locally, and look up current docs online if you need to.