The Data Scientist Show - Daliana Liu Bayesian thinking in work and life, ad attribution models and A/B testing, machine learning@Foursquare - Max Sklar - the data scientist show050
10 snips
Sep 13, 2022 AI Snips
Chapters
Transcript
Episode notes
Bayesian Inference Sparks Interest
- Max Sklar developed a deep interest in Bayesian inference during grad school due to its principled approach to updating beliefs with data.
- He values Bayesian methods because they allow clear encoding of prior beliefs and data-driven updates, akin to a computational scientific method.
Foursquare's Bayesian Rating System
- Max built Foursquare's venue rating system using user-generated tips as implicit sentiment data.
- He combined tips and explicit ratings for training and incorporated Bayesian priors to improve sparse and multilingual data modeling.
Use Bayesian Priors Prudently
- Always establish and continuously update your Bayesian priors, especially when lacking sufficient data for each item or feature.
- Priors act as regularization to prevent overfitting and enable principled decision-making as data accumulates.