I just found some notes that are more than a decade old on “stuff I want to learn”. It’s quite lovely to unearth that time capsule (ahem, looking at you, “logistic regression”, “survival analysis” and “numerical optimization”). I’m also surprised by how much the list is still on-point. And the terminology I was using back then is just different enough from the CS terminology I’ve been breathing for the past 8 years that it’s a good reminder of how many different disciplines have similar approaches for these concepts.

Just for grins, I updated my “wanna learn” list for 2023. Here’s where I still want more breadth and depth today:

- Statistics – deeper into framing ML as statistics and on experimental statistics (e.g., reasoning about error terms as in logit vs. probit models, revisiting the foundations of ANOVA)
- Time-series modeling
- Structural equation modeling
- Analysis (including metric vs. non-metric spaces)
- Abstract algebra (including group theory)
- A deeper course in linear algebra
- Signal processing
- Bayesian techniques (including hierarchical Bayesian modeling)
- Ecological inference
- A broader survey of probablistic algorithms (the family of LSH, Bloom filters, HyperLogLog++)

Now for the real trick – how do I bury this post so I can run into it in 10 more years and realize I’ve covered a third, a third is outdated, and a third would still be valuable to me….