(This article was contributed by the SP AISG Student Chapter)
When our student chapter was established, we were overwhelmed by the influx of new members coming from a diverse set of backgrounds. To get our members up to speed in terms of technical knowledge and ready to enter the realm of AI, we decided to organise a beginner-oriented bootcamp. Hence, SPAI’s Beginner Machine Learning Bootcamp was launched.
As we began planning, we soon realised that despite having prepared our learning materials, we were clueless about the flow and execution of our bootcamp. After seeking some advice from the Content Development team at AISG, we realised that our initial draft was too content heavy and the learning curve was too steep for a beginner.
Hence, we decided to take a step back and revamp our bootcamp by reducing the content and implementing several methods to smoothen the learning curve.
Smoothening the Learning Curve
One way we smoothened the learning curve was to break down and explain the intuition behind complex ideas through the use of metaphors and analogies.
For instance, when we introduce the concept of model validation in machine learning, we tapped into our audiences’ background by utilizing the analogy of an examination, to better establish the technique and intuition behind model validation.
Furthermore, we made the learning experience more engaging by incorporating interactive elements like quizzes and practical sessions. This was done in addition to frequent Q&A sessions for participants to clear their doubts.
Trial by Fire
After weeks of fine-tuning, we were finally satisfied with the curriculum that we had prepared. Then came our trial by fire: the actual run of the bootcamp.
The bootcamp turned out to be more successful than we anticipated, with a great deal of positive feedback from our participants. With that said, not everything went as well as we hoped (after all, no plan survives first contact with the enemy!), and several unforeseen challenges emerged during the event.
- To begin, we found that our curriculum structure was too linear as we had initially designed it such that each step in the data science workflow was taught in sequence. In practice, this structure made it difficult for some participants to comprehend certain concepts taught in the earlier days of the bootcamp as there was a lack of understanding of the bigger picture. For instance, without understanding how machine learning models train and work, how could one appreciate the benefits of data preprocessing like feature scaling? In retrospect, the bootcamp could have been further improved if we had started off more broadly with a look at an end to end machine learning project.
- Secondly, we discovered the pace of our code along sessions was too fast. To keep our participants engaged during the session, we removed large portions of our code examples and had our participants type out the missing code along with us as we explained the concept. Although it sounded good on paper, in practice, participants without much coding experience struggled to catch up. This was rectified in subsequent days, where we were more selective in which parts of the code to remove and found that this allowed participants to remain engaged and yet still be able to follow along.
- Finally, we found that the sessions were too long. Although we gave breaks at hourly intervals, participants still feedbacked that it was hard to stay focused in a 5-hour long webinar. Therefore, shorter sessions spread out across more days would better cater to our participants’ attention span.
The Journey Ahead
For the Operations team at SPAI, this bootcamp has provided us with invaluable experience, and we are thrilled that it has ended well. Moving forward, we hope to apply what we have learnt and continue delivering quality events for students and play our part to promote AI literacy within Singapore Polytechnic.
Oh Tien Cheng, Wong Zhao Wu
– SPAI Operation Committee
The views expressed in this article belong to the SPAI Operation Committee and may not represent those of AI Singapore.