Aaditya Ramdas – Checklists for Stat-ML PhD students

PhD students in Statstics and Machine Learning (and related areas) are expected to develop a fairly diverse suite of expertise by the time they graduate. For example, writing technical conference or journal papers, giving broad-audience or specialized talks, reviewing papers for conferences or journals, writing a research statement with a coherent vision, and so on. These are rather specialized skills which people have typically not received training on prior to grad school, but are expected to be excellent at by the time they finish a PhD.

We often expect students to develop these expertise without the “tripod” of directed mentorship, deliberate practice and constructive feedback. Currently, I think students partially succeed at this by some mixture of chance and personal motivation, and depending heavily on their advisor's own expertise on the aforementioned skills (we are not omnipotent!) and their advisor's motivation to provide mentorship. Further, many PhD programs do not have structured ways for students develop these skills.

The aim of the following checklists is primarily:

However, there are potential side-benefits:

Caveat 1: Ultimately, these are my opinions. If your opinions have a reasonably high correlation with mine, please feel free to share this page with others. If your opinions differ significantly from mine, perhaps make your own such page, so that students can make their own choices after reading multiple opinions.

Caveat 2: Such checklists are not meant to be complete in any sense (it would take too much time and effort). I typically use them to begin a discussion with my students in a group meeting (different topic each month, as needed), when I can answer a variety of related questions on each topic and add nuance to what I have written down. These are thus just beginner guides, or starting points, and entire booklets could be written on each.

Starting out:

In the middle:

At the end:

To be written: