ideas lab

| Feb 14, 2023 min read


I joined a research lab! I’m putting this up so it’s public, and that I can force myself to actually work on this so I will document what’s going on here


I had to do this a lot, and albeit not being the best, I think there are a quick few things that I think are worth mentioning.

Don’t expect to understand everything

It’s such a backwards way from how we do school reading. In school I expect to read something over and over and then it’ll make sense or I just go to OH and get it solved. Unfortunately in these situations you don’t have the time nor the luxury of someone just giving up time for you, so you have to be a bit more calculated with your efforts (still ask a prof / fellow research friends a bunch of questions, you still need that lmao). The best tips I got was to just break down papers into multiple passes.

Reading in Passes

First Pass: “What does it do?”

This one is pretty basic, just make sure that you have enough to give a 1-sentence summary. This part doesn’t have to be super drastic either, I like to just read the abstract, introduction, and the conclusion. The other important part is to just take notes of the headers of topic (even below the methods / related works, etc.) because it’ll start to get you subconciously noticing what the paper is about / what is in the domain.

To be honest, most papers I read will get this treatment (although I should be more thorough more often), so it’s important I can glance as much information as I can early on to decide (or really be told to) read more later on.

Second Pass: ELI5

This is where we start to actually pick up on the graphical information here. This includes stuff like figures, diagrams, illustrations, evaluation metrics, axes, etc. This stuff stores the most information because most of the paper is trying to explain what’s in these visuals (another reason why Data Viz is super important)

One thing I was afraid of earlier was reading the second pass and just not understanding much. It’s a dunning-kruger thing where you have to pop that bubble - you simply cannot gleam the whole paper from just the abstract, intuitition, and results - and just except that in order for this new information to make sense, we’ll just have to read some more.

Not knowing much

I’m really grateful that I’m not actually solo-ing my first real research project, I tried that before and without a strong domain background I really was struggling to succeed and not just be a weak link. Personally the best thing I could do is just be okay with asking a bunch of questions, in school you want to seem like someone who doesn’t take forever to grasp concepts but what I’m noticing is that it’s most time efficient and helpful to just let people know you have big gaps. It either means you’ll have more to read (and yes it’s also okay to ask questions when doing this) or you can save yourself a bunch of time down the road.

The other thing I should be doing more of is trying and failing. Whether it means writing that network, reading that blog post, just something outside my comfort zone. Research means we’re tackling unsolved problems, so it’s important that you start on train early on, saves hours down the road.