Supporting code for the Going beyond simple error analysis of ML systems blog post.

You can run each notebook in Colab or on your own. Just beware that Colab has a lot of the libraries pre-installed, but the versions can differ. For this reason I highly recommend running this line before launching all other code: pip install -r https://raw.githubusercontent.com/AlexandruBurlacu/error_analysis_code_samples/master/requirements.txt

1. L1
2. L2
3. L3
4. L4 (not done)

All examples have the NumPy random generator, the only one used, seeded. This means the code will be deterministic. If your results don't match what you see in notebook results in repo, please open an issue.