I am co-supervising a couple of research projects in ML/NLP for PhD and masters students. These students have decent programming skills and background through coursework but not a lot of research experience in the area. Developing a research project from a seed of an idea to something publication-worthy is a process that seems a bit opaque. A lot of this training involves observing more experienced students on their projects but it’s a very slow process and seems dependent on pure vibes right now. Both these students are fairly good at doing things when they are told what to do but are not making any progress coming up with ideas or developing next steps from an idea on their own. I want to find resources for them to refer to, that in some way pulls back the curtain on doing ML research. I know not all of it is teachable but things that might be good exemplars of problem formulation and following steps. I like Eamonn Keogh’s slides on a related topic, I am looking for slightly more drilled-down examples. Things like this research walkthrough or this old blogpost may be helpful I think.
Does anyone have any pointers or suggestions? Or even a suggestion for a different approach?
Comments
I generally tell the students what to do when they start. This is usually based on grants I have obtained or am in the process of submitting. It generally takes years to develop a good sense for research ideas, and I generally think it is better to get them going rather than spinning their wheels.