Why is modern architecture so ugly?

Apple’s M1 chips were released to much fanfare and some very impressive benchmarks. Unfortunately, all these gains don’t come for free. The downside of such a significant change is that it causes incompatibility issues. And because the basic CPU and instruction set are different, there are a lot of thorny issues. In this post, I’ll walk through some I ran into when installing standard data science libraries and how I solved them. In the end, I was able to get everything working, although it took quite a bit of patience. Hopefully, these incompatibilities get resolved relatively soon, but, until then, here’s my guide to help you out.

If you’ve ever tried to clone a repository from GitHub and gotten a “Permission denied (publickey)” error, you may need to create an ssh key and share it with GitHub. This post will walk through that process. The commands used in this post are for Mac and Linux.

This post is in a series on doing machine learning with unbalanced datasets. This post focuses on the evaluation aspect in particular. For background, please see the setup post.

This post is in a series on machine learning with unbalanced datasets. This post focuses on the makeup of the validation set in particular. For background, please see the setup post.

This post is in a series on machine learning with unbalanced datasets. This post focuses on the training aspect. For background, please see the setup post.

This post is the first in a series on working with unbalanced data. We’ll answer questions like how to train a model, how to validate it, and how to test it. Is it better than your datasets be balanced or representative of the real-world distribution?