The Duct Tape of Heroes Bayesian statistics
summary
Many concrete examples and code replace theory with results: a disease-test calculation that compares one versus multiple tests, a censored-exponential inference that recovers lambda from thresholded wait times, and an iterative KDE optimizer that bias-samples large triangles. Libraries pomegranate and daft and corner get demoed to build graphical models that handle missing inputs and update online. A half-million–game Heroes of the Storm dataset yields a Bayesian ensemble that claims to outperform random forests on imbalance detection.
AI-written summary · about
chapters (10)
- 0:00Bayesian stats in data science
- 1:05Bayes rule with dice and disease
- 5:36Bayesian inference with multiple tests
- 9:13Probabilistic models with Pomegranate
- 12:06Bayesian approach to biased data
- 20:31Bayesian optimization beyond DS
- 25:13Bayesian feature engineering in games
- 32:48Bayesian vs frequentist vs ML
- 36:41When to use Bayesian thinking
- 40:32Bayesian stats resources
related talks
39mThe Duct Tape of Heroes: Bayes Rule
17mByron Galbraith, Chief Data Scientist, Talla, NYC 2017
43mHow to look like a statistician (...)
42mBayes and Bool Walk into a Bar
34mFast, beautiful and easy Bayesian modelling: Can you have it all?
35mMental Models to Use and Avoid as a Data Scientist
25mScaling Bayesian Experimentation
01h 20mBayesian Inference for Fun and Profit