Example code accompanying the blog post Intro to Active Learning.

This notebook contains implementations of:

• Uncertainty Sampling with

  • least confidence
  • minimum margin
  • entropy criterion

• Query-by-Committee with

  • vote entropy
  • consensus entropy
  • maximum disagreement

• Expected Model Change for gradient-based learning

  • including a simple polynomial classification model

• Density weighting with KDEs

You can run this notebook in Jupyter Notebook (or Jupyter Lab), provided you have the following dependencies installed:

• numpy
• scipy
• matplotlib
• scikit-learn
• seaborn
• jax (only needed for expected model change)

You can also import this into a Colab, which comes with all these already installed.