What is APDKUPA

That is the abbreviation of "A post a day keep the unemployment away." I want to make fun of myself as unemployed since I am still looking for a job. During this period, I would like to write a blog every day just to consolidate the habit of writing. Considering the timing and the current job market, I guess I will have enough time to make it a habit.😊 The content might be a branch of coding implementation or a core concept explaination about certain ML topics. I hope my technical skills will be sharpened through these kinds of writable or speakable way.

Introduce myself

At this moment, I am a freshly graduated PhD from the University of Texas at Austin. My research interests are machine learning for climate change, more specifically, for building energy modeling and geospatial data. I came to the USA in 2021; before that, I was a master's student at National Taiwan University in the Civil Engineering department studying stochastic modeling and machine learning for precipitation.

Disclaimer

This series aims to talk about what I am interested in instead of what I am good at, although a large proportion of what I am interested in is probably what I am good at. I might make mistakes, probably a lot of them, so any discussion and correction are highly welcome!

Research vs. Engineering

That is the first topic that comes up when I am applying to jobs. Who am I and what is the best position that I represent? The answer is the ultimate safe card: it depends. I will further divide research into theoretical research and applied research. Let's take temperature prediction as an example. What I consider as engineering is taking a linear regression to predict the temperature for the next timestep. Using a known method on a task that has been proved to be effective is engineering to me. The hard part for these kinds of topics is to define research. I believe lots of people have experienced the struggle of research novelty.

Applied research is when you use some known methods that haven't been applied to this task. In our example, that would be using the Newest Energy-based State Space Machine to predict the temperature. Regardless of all kinds of follow-up concerns, it can be applied research where you bring something new to the table although the method and topic are known. In contrast, theoretical research is when you, based on certain theory—let's say primitive equations—develop a totally new method for this task, which hasn't existed in any existing works.

The vagueness is in how you treat the method. If I make some adjustments based on a linear method, does that make it applied research? What if I only change part of the model architecture, does that count as theoretical research? That is why I say "it depends!" at the start. The judgment is not made by me but by the reviewers you submit to. In my limited reviewer experience, the quality of the discussion section is how I decide whether this work is research. The work could be treated as good research if it has a very detailed explanation about the result, and discusses deeply the potential reasons for how it happened.

I am also eager to know how to differentiate theoretical and applied research since that is my "identity crisis." I am pretty clear that I am more of an applied scientist, but the positions that I want might require a theoretical science background, whether in climate science or machine learning. To me, the current answer is how much mathematics is involved in your work. If you have enough mathematics to back up your work, could it be considered theory? I don't know.

Conclusion

Science is open and has no destination. My PhD journey ends at a really weird spot where I have a feeling that I finally learned how to do research, but it's also okay since if that journey continues, then it doesn't matter where the node is.