blog

Blog

Three kinds of posts: implementation breakdowns of methods I actually run, paper summaries that compress a dense result into a five-minute read, and case studies on deploying ML against real industrial processes.

1 post

Tutorial

Offline Policy Evaluation Before You Touch the Plant

You cannot A/B test a water treatment plant. Here is how to decide whether an RL policy is worth deploying using only logged data from the incumbent controller.

  • Reinforcement Learning
  • Offline RL
  • Evaluation
  • Control

RSS feed