Conferences and Scientific Paper Sources

2023 May 21

Conferences and Scientific Paper Sources

A curiosity I always had during my student days was how professors stayed up to date with the state of the art in the literature where they were both writers and readers.

During my journey asking many masters and doctors around the world, I concluded that each researcher has a specific and unique method that meets various criteria of time, disposition, access, academic agenda, and even life stage.

But why stay updated?

I use my small channel ML Articles as a way to refine my knowledge about a subject I have a genuine intellectual curiosity to learn more about. This gives me a lot of freedom to think about things outside the news or hype cycle.

Not to mention that I don’t have any kind of academic agenda to fulfill, which I particularly like and find intellectually liberating, as I can let my curiosity guide me through a less structured self-learning process.

Although it is unusual for someone in the industry, I like to stay updated through academic literature for a number of reasons:

Intellectual straitjacket of the corporate world: Since I am in the corporate world, learning through practice often becomes a limiting factor in my understanding process from the perspective of experimentation rigidity. There are few places where mistakes are seen as learning rather than loss. A large part of scientific works does not have this limitation as severely because there is freedom to understand theory and apply practice in a systematized and scientific way;

Leaving the corporate off-road and hitting the Formula 1 track methodologically halfway through the race: Depending on the research, an initial path to a solution has already been created and validated systematically, exposing potential and limitations.

I personally don’t like following the results of the papers (due to confirmation, congruence, observer-expectancy, decoy effect, selective perception, and Semmelweis reflex biases) but rather the methods these papers bring and how I can apply them in my day-to-day.

Thanks to this form of prior exploration, I started using some techniques like Bag of Tricks in NLP, scalable boosting methods back in late 2016, and computer vision convolution architecture with convolution block separation in 2017;

The counterpart to this in the corporate world is all the biases described above, mixed in a swamp of nonsense, various corporate pressures, and the erroneous use of instruments like A/B tests in a bunch of senseless experiments driven only by bosses’ egos.

  1. It is a type of literature that puts me in contact with the state of the art, the potential, and the limitations of my field of knowledge; this contact helps me be a better-informed professional regarding practice and theory;

  2. Depending on the degree of maturity of a certain area, I can have a general overview of how to navigate professionally within my field of interest in a more conscious and informed way.

Here in item 4 is where I see many people getting confused. I know Data Scientists and ML Engineers who are very talented getting “bored” with something specific within an ML field and then jumping to something totally unrelated, like “front-end” or “management,” which are useful skills in the modern job market but have a low degree of transitivity between them.

I want to write about this one day, but in short, having a bunch of unrelated qualities will only dilute the time to achieve mastery in a topic (which at the end of the day is the fortress the professional needs) and will not generate the compound effect (or snowball) where underlying skills build a long-term career.

What to read?

This is the most important question of all, because time is finite and thousands of papers are published every day.

I don’t have a simple answer for this, given that it depends on each person’s academic and professional moment, but what I have for myself is a criterion where I consider two dimensions which are (i) the intensity of the moment (could be professional, personal, or both) and (ii) the need for reading.

a) Low intensity without the need for critical reading. Here I practice intellectual indulgence proper. I read what interests me without any specific criteria. Some readings range from bilingual development of children in early childhood and the cognitive issues involved, how most scientific medical findings are false, and how general venting and complaining doesn’t work and how this is pushing people into states of permanent dissatisfaction and even aggressiveness.

b) High Intensity without the need for critical reading. Here I take a break from readings that demand more reflection, but I’m always looking at something more recent just to “know what’s happening,” but nothing that will take my focus off the current moment.

c) Low Intensity with the need for critical reading. This state is when I know some interesting project is coming on the horizon that will require either an update on something I’ve been practicing for a long time or if I’m going to lead some collective implementation effort which will require me to communicate with multiple people. An example was when I had to work on a computer vision project where I had to meet algorithm performance requirements (i.e., Recall at top@5 above 90%) and system performance (i.e., response time below 50ms), where I had to read many papers to reach something satisfactory and explain it to data scientists + PO + SysAdmins + Operations with the rationale for the choice.

d) High Intensity with the need for critical reading: In this type of situation, I take a break from some activities.

Where to read?

Some people read better at work to make better use of professional time, some like to read in waiting situations that I call buffer, some have a specific place to read like their own personal office or outdoors.

Personally, I like to use waiting situations for my readings.

And anything goes here: Uber rides, airport waiting rooms, flights, buses. These situations are usually “dead” moments where most of the time socialization won’t happen anyway, and time will be lost regardless; so I try to capitalize on these moments.

However, I have a specific day of the week when I do my readings and it’s something that for me I try to maintain an almost religious consistency. Rain or shine, I will read on that day of the week, even if it’s for only 25 minutes. I’ve done readings in police station rooms, wedding warm-up parties, maternity waiting rooms, during funeral vigils, in the middle of the beach, and at kids’ birthday parties!

My inspiration for this came entirely from Jerry Seinfeld and his method called “Don’t break the chain”.

This helps me create a habit in a simple way, and maintaining discipline is very easy. Regardless of my state of mind or disposition, I will read on that day of the week.

I definitely do not recommend such a strict agenda because this requires very high family and social understanding depending on the relationship.

Conferences

I have a very clear separation of conferences into levels. And here I don’t put it in qualitative terms per se, but by a degree of priority in terms of reading and monitoring. I don’t follow the H-index or any kind of pre-constructed list by the academic community, due to the fact that I have distinct interests.

Currently, I have 3 levels:

Level 1 Events whose dates I actively follow and the moment the proceedings are released, I already try to read as fast as possible if it’s related to work or some personal research. In most cases, they will be journals with more applied works.

Level 2 These events are those that have relatively high relevance in the mainstream and have good editorial lines. I try to follow them as much as possible.

Level 3 Here are all the others where I only research something truly actionable or some interesting idea that ended up being missed in the levels above. There are some journals from adjacent or theoretical areas.

It’s important to point out that these are very personal criteria and it took me a while to reach them.

An example of this is that obviously NeurIPS and CVPR are the best conferences in Deep Learning and Computer Vision, but in my case, most of the works that come out there, by my criteria, are much more informational (SOTA, etc.) than something I can use in my day-to-day like RecSys, where every edition I have a ton of papers to review because it directly impacts my day-to-day.

And market events?

With the technical emptying of InfoQ events and the end of O’Reilly events, the market events that most interest me are those from the Linux Foundation; especially Kubecon.

But personally, I confess that fewer and fewer events catch my attention every day, but that’s a subject for another post.