MSc Data Science, how about it?
What an MSc Data Science student learns and how it compares to online learning
Throughout the year, people have messaged me and my course mates about how we have found the MSc Data Science at the University of Bath. Now that the taught section of the course has finished, I thought it would be a good idea to write a blog post about the experience to help people decide whether the course is for them.
What this article covers:
- Why I chose to study an MSc Data Science
- What did I learn?
- Comparison to online courses
Why I chose to study an MSc Data Science
I had just finished my undergraduate degree in Economics and I was doing some research with a couple of my lecturers over the summer. I had always found inequality an interesting topic in Economics so when I found out that my lecturer was researching how AI and automation was affecting inequality, I jumped at the opportunity to get some hands on research experience.
This research experience completely changed my perspective on what I wanted to do. I realised that I wanted to be part of the innovation and growth taking place, as opposed to studying it from the sidelines. Although I believe that studying social issues is necessary, learning about the effects of AI and automation inspired me to get involved in the action. After realising I needed to update my skill set, I decided to enrol in the Python for Everybody specialisation by Dr Chuck on Coursera (which I highly recommend as an introduction to Python / coding).
After getting familiar with Python, it was only a matter of time before I stumbled across the mystery world of data science. I had always enjoyed econometrics, the application of statistics to economics, so data science sounded even cooler. Due to my econometric mindset, my first impression of data science and machine learning was ‘well, this is complete madness.’ Curious about the madness, the hype and the potential for innovation and economic growth, I continued learning more about data science until I decided to pursue a masters.
One important consideration for everyone studying a masters is the financial burden. In the UK, we are fortunate enough to have very reasonable student loans. However, the only reason I could afford the masters was because I received a scholarship. This highlights a large source of funding that most students assume they won’t receive. If you’re interested in studying any masters, I strongly encourage you to apply for every scholarship you can find — even if you’re not eligible for it! Universities have cash allocated for this so use it!
Takeaway: I wanted to switch from watching on the sidelines (economics) to getting my hands dirty and adding value to society. The madness and hype around data science made me wonder if it was just hype. The masters was made affordable for me by the vast amount of scholarships available to students.
What did I learn?
There were some specific skills that I wanted to gain from studying the course. Specifically, I wanted to learn to code, to handle big data, to produce clear and effective visualisations to portray a story, and to build custom data sets. The course delivered on all these fronts by covering Python, SQL, design principles for effective data visualisations, and data collection techniques.
Of course, no data science course would be complete without machine learning. Machine learning, and more specifically deep learning, can often be seen as the pinnacle of data science that attracts all the media attention. So what did I learn? I learnt the fundamentals of machine learning through various coursework that involved coding algorithms, such as random forests and AdaBoost, from scratch. I also used neural networks for deep reinforcement learning and image classification, and I will most likely use deep learning for my research project this summer.
Although deep learning is not covered extensively in the course, you get very familiar with the fundamentals of machine learning and break out into the world of Bayesian machine learning as well. Personally, I enjoyed this set up because it helped me establish good practices before diving into deep learning. However, I understand that in the age of implementation, a lot of machine learning applications involve using transfer learning from a pre-trained state of the art model.
When transitioning to a new field, it can be easy to get caught up in all the cool stuff that you’ve learnt. I certainly felt this way when I first learnt to code — it felt like I had a super power. However, when thinking about what you have learnt, it is also important to keep in mind what you haven’t learnt. If like me you haven’t worked in industry, it can be hard to think of what the key skills you’re missing are. By taking advantage of some great data science content creators (e.g. Ken Jee and Vin Vashishta), I quickly developed a list of skills to work on. Some of these include: the ability to understand whether a business problem is machine learnable, moving beyond the safety of Jupyter notebooks to deploy models, and really focusing on the value added by data. If you can work on these skills while you’re first getting into data science, you’ll have a great advantage over the competition.
Takeaway: I learnt a lot and got my hands dirty with some interesting projects. If you pursue a masters, keep in mind what skills you’re not gaining and consider if its worth investing some time into developing them.
Comparison to online courses
A lot of eager people wanting to gain data skills have resorted to online and self learning as an alternative to pursuing a traditional masters. Daniel Bourke’s self-created AI masters is a great one that springs to mind. So which is better? I’m not going to pretend that I know the answer to this question but I will provide some insights that I’ve gained during my formal study.
Projects
If you’re thinking about learning data science, chances are that you want to get into the data industry. The best way to land a job is to show off what projects you’ve worked on. I was pleasantly surprised when I realised how project based the masters course was. Going from virtually all exams during my undergraduate to virtually all coursework in the masters was great. Now I can talk about the projects I’ve done instead of just saying: look at me, I got x%.
It is also worth noting that personal projects are even better than coursework as they show that you’re able to apply what you’ve learnt to a problem that you’ve identified. The important part here is to demonstrate your ability to create value for businesses. I’ve just started my first personal project and had to scrap a load of ideas because I didn’t feel they created value. Now I am focusing on a project to help drug discovery researchers using PubMed data, therefore creating real value.
Whether you are brushing up on some skills or learning concepts from scratch, online courses can be a great resource. However, there is a real temptation to fly through as many as you can at x1.5 speed and not apply anything you’ve learnt. With the masters, this isn’t really an option. The set up forces you to complete coursework after coursework using the lectures as guidance, as opposed to the lectures being the main focus. In my opinion, this was the biggest strength of the masters. You had to figure out how to solve a problem. It didn’t matter whether you used lecture content or not, it just needed to be solved.
Side note: Some people have asked me what kind of projects I did during the year so here is a list of some notable ones: sentiment analysis, image classification (deep learning and other techniques), regression (continuous variable prediction) on crop yield data, providing investment recommendations for electricity generation using time series forecast, producing infographics, working with clean and unclean data, solving the lunar lander environment by Open AI using deep reinforcement learning.
Cost
If you compare the price of a monthly subscription of an online course to the cost of a masters, this is a no-brainer. Online learning is far cheaper. However, as I mentioned at the beginning, if you are eager to pursue a masters but can’t afford it, apply for every scholarship you can find. This is definitely a source of funding that students overlook. I did too. I always thought a masters would be great but there’s no way I could afford it. A few scholarship applications later and I was set. This drastically reduces the cost of a masters. Bare this in mind if you’re choosing between the traditional masters or online learning for any subject.
Network
They say its who you know, not what you know. Although it’s definitely possible, it’s much harder to create a network of like-minded people through online courses. Most courses nowadays have an online community that you can connect with on either slack or discord. However, you have to be proactive to get to know people in these communities. In contrast, universities are network creators by design. For example, the computer science society at my university hosted a hackathon and within minutes of meeting my team, they were happy to refer me for a data science role. Networking works. If you’re going the online route, look for online or local hackathons or meet-ups to meet people.
Content
Online course providers and universities have different motives. Ultimately, universities are training you to be academics. As a result, you tend to go more in depth on the theory behind the techniques you use. However, the beauty of online courses is that you can tailor them to your liking. If you want more technical content, go for it.
This highlights the fact that online courses are more flexible than universities. If you choose formal study, make sure you’ve done your research on the course content. The advantage of the traditional masters is that it’s easier to lose motivation and go round in circles when studying online. The structure of the masters I studied was constructed by two very strong academics so I felt confident that I was studying the right content to gain a good understanding of data science. Obviously, the content covered by a masters will vary from course to course but the research project is one area where you get freedom to work on what you want, provided you have someone to supervise you.
Although it might sound like I favour the traditional masters in terms of content, most lecturers actively encourage students to use additional resources. With data science and machine learning being very open communities, most of these resources are freely available online. All in all, there really isn’t that much difference in terms of content, provided that you plan your studying well.
Takeaway: Online study and the traditional masters both have their pros and cons but its down to personal preference on which is best for you. Do your research and apply for scholarships if you do a masters!
Conclusion
Overall, I’ve thoroughly enjoyed the course and would recommend it to most people. Becoming more data literate has changed the way I see things day-to-day. Most things can be related back to data in some way or another. The biggest benefit that I’ve noticed from studying data science is my shift in attitude towards skill acquisition. Nowadays, I’m confident that I can pick up a new skill quick enough for it to provide value. But the masters is just the start of a continual learning journey and I’m looking forward to seeing how the rest of it plays out.
Thanks for reading!
If you have any questions about the course or otherwise, feel free to connect with me on LinkedIn and send me a message!