PermGrad v1.0.0 – Official Release

This release contains the full code, experiments, synthetic-image generation pipelines, and interpretability analyses used in the manuscript:

"PermGrad: A Hybrid Neural Network Framework for Unified Symbolic–Spatial Interpretability in Tabular Deep Learning"

Included in this release:

• Training scripts for MLP, CNN, and HyNN architectures (classification and regression)
• TINTO-based synthetic image generation workflow
• Permutation importance, Grad-CAM, and PermGrad interpretability modules
• All experiment configurations and reproducible settings
• Graphical abstract and figures used in the paper

This version corresponds to the exact codebase used for the submission to Pattern Recognition – Special Issue on Evolving Multi-View Learning.