Three Lessons I learnt during my Machine Learning degree

Last June, I was overjoyed when I got accepted into the MSc Machine Learning program at UCL. Pursuing this degree was my aspiration every since the second year of my undergraduate studies. Although I must confess, making the shift from a pure life science subject to a quantitative one, I was quite nervous and didn’t know what to expect. Having completed six months in the course now, it has been a steep learning curve for me, but looking back on where I was a year back, I feel proud of how much I have accomplished and excited about what more I will learn. So with that, I wanted to share my experiences over these past months.

What have my days been like?

Given the current circumstances, my entire degree has been virtual, which is a bit unfortunate as I couldn’t travel back to London, attend university, and make friends in person. But overall, I would say it still has been enjoyable. The lectures have been super organized, informative and interactive as well. What I found challenging, however, was adjusting to the time differences, being eight hours ahead in Malaysia. Most sessions would be recorded to access them at my own time, but I preferred to watch them live, as it would allow me to interact with and ask questions to my professors. This required me to work during the late evenings, and some classes even ran between 12 am – 2 am! It wasn’t easy to sit through these sessions and concentrate initially, but I guess I got used to it over time. Thanks to platforms like Discord, I also made many friends (virtually) who live in similar time-zones as myself, and we pulled through together.

Now, in terms of the workload, the course has been incredibly rigorous. Almost every two weeks, I’ve had to face a new assignment, unlike my medical science degree, where I would get a single project that lasted nearly three months. Since the projects were all programming-based, they were time-consuming, as 1) I was implementing complex models like support vector machines and neural networks from scratch, 2) there was a lot of debugging involved, and 3) there was an added pressure of making the programs computationally efficient; otherwise results wouldn’t come in time. The process was laborious, but I can confidently say that I came out intellectually fulfilled by the end of every assignment! Also, many of the projects were team-based, so it was even more fun to work with my friends and engage in exciting team discussions.

My three most significant learnings from pursuing an ML degree

Having reached the halfway point in this course, I feel much more confident and knowledgeable about working in the machine learning field. Below are some of my biggest takeaways from this experience. These are just my personal opinions, and I’m sure others will have different thoughts.

1. In ML, it’s essential to be a quick learner. Being a reasonably theoretical course, I’ve learned that the amount of depth to which one can explore a specific topic is endless. ML is an amalgamation of so many fields like Statistical learning, Numerical Optimization, Information theory, and Algorithmics, for which people have written hundreds of treatises. Reflecting on my study methods before this course,  being a very detail-oriented person, my first move on seeing a challenging concept was searching for a full lecture-series on YouTube that explains the A-Z of the topic to understand that concept well. But over the first few weeks, it started to dawn upon me that the luxury of time is not available during a fast-paced course like this. To study the content effectively, I had to modify my learning strategy by staying away from the long 2-hour long YouTube lectures and looking for the short 10-minute tutorials. So far, this method has worked out well for me. A valuable piece of advice I also got from one of my professors in the course was that when applying ML in industry, it is more important to get an intuition of the ideas when studying literature rather than getting absorbed in the math. So, in short, be a quick learner whose focus is on the bigger ideas rather than the nitty-gritty.
2. Want to pursue a career in ML? Know what domain you want to work in before you step in! Having completed my undergraduate studies in medical sciences, I’ve always been passionate about combining ML methods in the same domain. However, from interactions with my course mates, I noticed that some friends couldn’t initially decide which area to apply their ML skills. Especially with the research project hunting season begun, there has been so much choice for project topics that making a final decision can be challenging. Especially when it comes to career searching, it can get tough, considering that ML has wide-ranging applications in many domains. Thus, another learning I wanted to share is that it’s always best to discover your domain interest before considering the type of ML you will apply. Doing so can help streamline the search for your optimal career.
3. Being a good scientific report writer can take you a long way. Remembering the coursework I did in my medical sciences degree, most of it was essay-based, where I would review previously conducted research. Whereas now, most of my projects are involved – I design and execute my own experiments and then produce a report where I discuss the methods/theory implemented, create visualizations of the results, and discuss the same in more depth. From my understanding of the ML field, most jobs in the industrial and academic setting are likely to be research-based, where one would have to produce a research paper or a brief report like this. Hence, developing the skill of articulating oneself and presenting a crisp and concrete analysis is one that I found to be extremely valuable through my modules. Thus, I would say that for anyone interested in studying ML formally, it would be beneficial to practice this scientific writing skill, especially using a typesetting system like Latex.