Active Learning Demo

This is a basic script I used for showcasing an Active Learning approach using Gaussian Processes (GP) for a presentation on improved sampling strategies for computationally expensive functions.

Notes

• The actual Active Learning part is extremely simple, the next sampling point is selected based solely on the maximum standard deviation of the GP.
• In previous roles I have implemented more robust approaches that better balance exploration versus exploitation such as using maximum entropy or the Mismatch-first farthest-traversal heuristic to select the next sampling point.
• Included in this repo is a simple graphic I made using imagemagick and the outputs of the active_learning.py script.