Contextual Bandits Research

As of last week, I’ve started doing research with an incredible professor at my university! His work revolves around using machine learning to develop personalized online education systems, something I’ve dreamed of developing since my teen years. These systems factor in a user’s background (familiarity with course context, self-professed learning style, etc.) when providing explanations for new topics and problems, facilitating a deeper contextual understanding.

For instance, let’s use the situation of two sets of students taking an introductory programming course: students new to coding and students experienced at coding. While the experienced coders might find text explanations more helpful, it might be the case that those new to coding prefer videos. Personalized learning systems take this context into account and adapt future explanations appropriately. In this way, learners are given the explanations that are best for them.

What if there’s a subset of experienced coders that are actually audio-visual learners? Though on average text-based explanations might be more helpful for experienced coders, certainly videos would work better for these students. Given enough data, these systems can discern the existence of such subgroups and modify their learning experiences accordingly. In essence, personalized learning systems provide individually-tailored educational experiences, and on account of their machine learning base, the more they are used, the better they become.

For now, I’m still trying to learn the statistical concepts underlying the technology, but I hope to improve my skills over the coming weeks and participate more in this cutting-edge research!

First Glimpse at Computer Vision

My first competition in the Kaggle community is the Digit Recognizer competition run by Kaggle. The competition’s objective is to teach Kagglers the fundamentals of computer vision, and as such is part tutorial, part competition.

I’m extremely excited about this competition for multiple reasons. First and foremost, I’m finally going to apply the ML fundamentals I picked up from Kaggle Learn over the last week. I wasn’t sure about how to start this process from scratch, so the format of this competition suits my needs perfectly. Second, computer vision is a ridiculously popular subfield of AI! Tech giants across the board are investing heavily in advanced computer vision research, and many of Kaggle’s top-tier competitions involve image classification in some form. It’s pretty satisfying knowing that the skills I’ll develop from this competition have immediate and pressing applications.

Most of all, I’m excited for this competition because it represents the first step of my journey towards becoming a real data scientist. It’s the confidence that I’ll gain from self-improvement that will ultimately have the biggest impact going forwards, and I think this competition serves as the perfect medium for that to happen. On that note, here goes nothing!