✨ Explain any ML models anywhere with WebSHAP + ONNX Web Runtime ✨

[xiaohk]

Hi all,

Thank you so much for creating and maintaining ONNX Runtime!

I'd love to share our recent work WebSHAP, a JavaScript library that explains any ML models completely in users' Web browsers by computing SHAP values! In other words, users can explain ML models directly on their phones, TVs, and even smart refrigerators!

We provide three demo examples that explain XGBoost, CNNs, transformer classifiers on our demo website. For two examples, we used ONNX Web Runtime. I was very impressed to see how easy it was to use WebSHAP with ONNX: users only had to wrap their ONNX inference code into a function, and everything just worked :)

Thank again for your work on ONNX Runtime. It's a game changer for making ML accessible anywhere! ❤️ Please let me know if you have any question!

🚀 Demo: https://poloclub.github.io/webshap/
🐙 Code: https://github.com/poloclub/webshap
📘 Research Paper: https://arxiv.org/abs/2303.09545