[P] A new open source MLP symbolic distillation and analysis tool Project

[P] Hey folks! I built a tool that turns neural networks into readable math formulas - SDHCE I've been working on a small project called SDHCE (Symbolic Distillation via Hierarchical Concept Extraction) and wanted to share it here. The core idea: after you train a neural network, SDHCE extracts a human-readable concept hierarchy directly from the weights - no extra data needed. It then checks whether that hierarchy alone can reproduce the network's predictions. If it can, you get a compact symbolic formula at the end that you could implement by hand and throw the network away.