Contribute Media
A thank you to everyone who makes this possible: Read More

The Duct Tape of Heroes Bayesian statistics

Description

PyData London 2016

In this talk I will give many examples of when Bayes rule will help you in your day to day work. I'll quickly show many examples of bayesian statistical thinking in action; the pleasure of inference, probabilistic graphs, model selection, feature generation, even operations research! I'll finish with a dataset from Heroes of the Storm and I'll show why Bayesian models can outperform randomforests.

My talk is made up of the following examples;

  • basic disease example: what is the value of adding an extra test to a patient
  • give an example of an inference task that is very hard to do properly without bayesian thinking
  • creating simple probibalistic models with pandas and showing how they are robust against missing data
  • demo the daft, corner and pomegrenate library
  • show how you can use bayes rule to pick models
  • demo a bayesian probablistic approach to finding overpowered characters in the Heroes of the Storm video game.

Slides available here: http://koaning.io/theme/notebooks/bayes.pdf

Details

Improve this page