Shapley Values Explained | Interpretability for AI models, even LLMs!

Ever wondered how to interpret your machine learning models? 🤔 We explain a powerful interpretability technique for machine learning models: Shapley Values. They can be used to explain any model. 💻 We show a simple example code of how they work, and then 📖 explain the theory behind them. AssemblyAI (Sponsor) 👉 https://www.assemblyai.com/research/universal-1/?utm_source=youtube&utm_medium=social&utm_campaign=universal1_letitia AI Coffee Break Merch! 🛍️ https://ai-coffee-break-with-letitia.myspreadshop.de/ Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏 Dres. Trost GbR, Siltax, Vignesh Valliappan, Michael, Sunny Dhiana, Andy Ma Outline: 00:00 Interpretability in AI 01:02 AssemblyAI (Sponsor) 02:23 Simple example 03:51 Code example: SHAP 05:17 Shapley Values explained 07:59 Shortcomings of Shapley Values 💻 Demo for SHAP on LLaMA 2 LLM: https://drive.google.com/drive/folders/1EE2F5fbrBMO28DWzcKll9V_MImydsg3N?usp=sharing Keep in mind that you need to have the resources to run LLaMA 2. If not, try out the “gpt2” model in the code. You can find simple examples here: https://shap.readthedocs.io/en/latest/ (see e.g., “Text examples”) 📙“Interpretable Machine Learning” by C. Molnar: https://christophm.github.io/interpretable-ml-book/ ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕ Patreon: https://www.patreon.com/AICoffeeBreak Ko-fi: https://ko-fi.com/aicoffeebreak Join this channel to get access to perks: https://www.youtube.com/channel/UCobqgqE4i5Kf7wrxRxhToQA/join ▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀ 🔗 Links: AICoffeeBreakQuiz: https://www.youtube.com/c/AICoffeeBreak/community Twitter: https://twitter.com/AICoffeeBreak Reddit: https://www.reddit.com/r/AICoffeeBreak/ YouTube: https://www.youtube.com/AICoffeeBreak #AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​ Video editing: Nils Trost