Found the following on my drafts while writing a post about finishing my masters, posting it here despite the delay for posterity.
I’m writing this in May of 2020. The world is in the middle (beginning haha) of a pandemic. Brazil, as always, is in the middle of many crises (political, sanitary, and economic). And at the same time, I’m approaching the middle of my masters, time to reflect on what happened and plan for the future.
One of my goals for 2020 was to put myself in new situations, I had no idea of what I wanted for my professional career. One path that seemed interesting was an academic career: stability, endless learning opportunities, exploring cutting edge topics, sharing knowledge. Still, how to know if this was a fit for me? The answer, I applied and got approved to a master’s program, with any luck two years would be enough to understand what it means to be a researcher.
I should also point here that I have a full-time job as a Software Engineer, which is also challenging and fun (sometimes). I will not get into details of my work here, know that I like it, amazing people to learn from, amazing company, etc.
What does it means to do a masters in Brazil
I know some countries require one extra year after graduation to earn the master’s title, here we do a bit differently. A master’s degree in Brazil takes 2 years. The default path is to take one year of classes and one year to research. In a normal situation, I would be finishing classes and preparing to focus exclusively on research, however, due to COVID-19 all the classes and university activities are suspended, alongside all non-essential services.
In Brazil, you can earn a masters by attending a private institution, however, the best opportunities are in public universities. After a lengthy admission process, I was accepted into the Computer Science program of the University of Pernambuco, that’s to say I don’t pay for the program myself.
My Research
I’m part of a research group and our goal is to use gait information and artificial intelligence to solve health problems. If you are confused about what gait means, it means walking, an automatic behavior that does not demand cognitive effort. You may also be confused about the utility of gait information and I can tell you there are numerous. For example, many researchers point to gait as a unique identifiable part of an individual which means that there are no persons with the same gait. Others use gait information to classify age and gender. Gait analysis is also used to track the progression of diseases or even monitor recovery. My research focuses on using gait information to detect early signals of neurodegenerative diseases. Identifying a neurodegenerative disease early has the potential to improve the quality of life of the patient giving the physician more treatment time and hopefully delaying the impairments caused by diseases.
How I approach it
A decision I made at admission time was not to push to be the best. This is an experiment, the goal is to know if this can be a viable career path. In other words, I wanted to do good, give it 80% and save some energy to grow in other environments, do not get exhausted, do not get burnout, still have a life. To assure physical and mental health I was forced to organize myself better, here’s how:
Show up to work
I made the commitment to work at least one hour per day on my research, on good and bad days. In the last three months, I have skipped eight days. The problem with daily commitments is that just showing up may not be enough, you can be distracted, you can be tired, you can be sleepy, it’s paramount to define what do you want to accomplish in those hours. For me, it was difficult to measure research accomplishments, especially while doing a systematic review (I will get back to this in the next topic), so my way of making sure I was progressing was tracking the hour of deep work and making sure that I had enough hours in the week. A minor detail: studying for classes doesn’t count as research hours.
Why one hour? With all the day-to-day commitments: work, classes, commute (don’t forget traffic), one hour was what I could guarantee, it was something I knew I could achieve. On the other side being completely honest, I have other interests.
Stick with the boring stuff
The most valuable thing to do when starting research is to do a systematic review, you must know your research field before proposing anything new. My first step was a mini systematic review (“mini” because I’m a master’s student with a time constraint of two years). Selecting the right articles was already a fight on its own, my co-advisor insisted on a protocol to perform the analysis of the articles, which led me to endless readings on methods to handle systematic reviews. It was not the most exciting thing to do a the beginning of my research, but I stick with it. The result was 113 articles found in many databases filtered by keywords. From the previous total, 59 articles were filtered by title, and 43 articles after a second filter by abstract. These selected articles would consist of my “mini” systematic review.
Searching for more
I finished the articles from the systematic review two weeks after my planned stipulated date. I assumed that after reading the articles I would have a clear picture of my field and what contributions I could give, but the truth was that I had more questions than answers, which triggered a new round of paper selection. With a lot of notes and articles under my belt, I was able to distinguish good papers from bad papers and just dig deeper into the good ones, following their references and checking their citations, as result, many other great works appeared to complement my research.
After some time, I noticed that good papers quoted almost the same group of authors. The good authors by themselves only worked in a very specific domain. I saw myself reading papers that were very similar in methods and results, it was like every paper was a refinement of the previous one. Around this period I was reading a book called Range, about how generalization is much more useful than specialization (I enjoyed this book very much). In chapter 8 the book discusses the big push of society for hyper-specialization and how this creates a gap for non-experts. Bare with me for a bit. Experts have built an amazing core of knowledge about a specific field, however, when presented with a problem, they tend to do a local search, many times losing a globally optimal solution because is not in their area of expertise, this can also be called the Einstellung effect. Non-experts may thrive in these situations by coming up with analogies from different fields to solve a problem, in many ways outside knowledge may be the key. Swanson, a researcher, put this to test, he tried to make discoveries by connecting information from scientific articles in subspecialty domains that never cited one another and that had no scientists who worked together. By systematically cross-referencing databases of literature from different disciplines, he uncovered “eleven neglected connections” between magnesium deficiency and migraine research, and proposed that they be tested. Swanson showed that areas of specialist literature that never normally overlapped were rife with hidden interdisciplinary treasures waiting to be connected. Creating an undiscovered public knowledge database just waiting for the connections to be made.
Back to my research, the thought of having answers to the problems of my field documented and waiting to be discovered moved me. As a result, I started to explore solutions outside my field. I did it on my own and it’s unknown to me if this practice is good or bad. I can say that my co-advisor was strict about following the systematic review methodology, which I followed at the beginning but abandoned at this point. To explore solutions from different fields I used a tool called Arrowsmith which identifies meaningful links between two sets of articles. It works by performing two searches to define “literature A” and “literature C” and then generates a “B-list” of words and phrases common in both works of literature. The B-list is displayed ranked by relevance and can be restricted to certain semantic categories. The key is always the keywords you use to define the literature, in my case if I used the same keywords from my systematic review, as a result, the generated articles were almost the same ones from my traditional systematic review.
My research is all about feature extraction and classification. Neurodegenerative diseases however common they are, are still the exception, it sounds like another way to describe the problem would be anomaly detection or even outlier detection. I asked myself how to rephrase the problem I had to find connections between different fields and new articles started to appear. I found myself reading papers about Online Ship Prediction, Anomaly Detection in Industrial Wireless Networks, and Outlier Detection. The papers started to look new again and their approaches to problem-solving exciting. I developed a particular interest in complex systems and the theory behind them.
Where I’m now
The time to review the literature is over. Going forward I need to focus on my thesis and my proposal. The reviews allowed me to create a robust background to decide what are the explored areas and where I should concentrate my efforts, alongside with dozen of ideas. In the next months, I will be presenting my Thesis Proposal to the university board.
Lastly, after one year, I come to notice that most of the things I did, in the beginning, were useful to build my researcher skillset. I didn’t know how to read articles, how to take notes, and how to evaluate the quality of the articles.