A few years ago, in the middle of Covid-19 pandemic, I took up photography. It had been ten years since I held a DSLR. Camera technology had drastically changed during that time. The industry was moving towards mirrorless. I desperately need a hobby and I chose to learn this technology to rekindle my old passion that is photography. Three years and ten thousand pictures later, I feel content. I learnt a lot about taking photos. I traveled occasionally, shot nice pictures, learnt to edit, and publish. I even built a website to host my creations. More than anything, I cherished the joy of learning and creating.
I’m wondering if I could choose a new area to focus my learning. How about AI and Deep Learning? Well, I am aware that photography and Artificial Intelligence are not the same. While I have had some formal introduction to data science in my graduate school and have built predictive models in the past, I have not done any serious work in this area in the past few years. But in my day job, I validate risk models for a living. I am also a power user of R/RStudio (this website is built in Quarto) and am very comfortable with Python too. So, I think I am not miles away from feeling at home.
Why
ML/AI & DL have changed a lot since 2015. Their impact is felt everywhere (case in point: Large Language Models and their applications). Like it or not analytical models will impact all aspects of our life in meaningful ways. I wish to be more than a spectator. Good thing is, world is not stuck in 2015 when it comes to online resources. Then there is Kaggle where one could learn and compete. I am sure this is going to be an amazing journey! I also intend to record my progress and share my experience through this blog, something I wish I did during my photography journey.
What
So what’s my approach going to be? Tackle one dataset at a time. I will analyze Kaggle competitions to understand how top rank solutions were built. After racking up enough knowledge, I shall start competing. I wish to stay away from video centric MOOCs. I know this is not how everyone operates. But I believe in learning-while-doing. That does not mean the theory is taken for granted. I shall record all my model related learning in the concepts section of this website under “Analytics Modeling”.
Targets
Owing to the rapid evaluation of predictive and prescriptive analytics, any goal we set will need to be reevaluated in short time period. It is also impossible to scale this subject as an outsider. So I am approaching this task differently. I used the word ‘content’ at the start of this article when I spoke about my photography journey. I believe it holds true in this case as well. I do not want to miss the joy of learning and doing in pursuit of moving targets. Of course, validation is awesome. But I would rather work out till I feel the pain!
Next-Steps
This article will expand. I shall link other articles and analysis to this page. To start with I will discuss my compute set-up and resources.
Compute & Set-up
The first order of business is to set up the rig. This article here explains the decision process I employed in arriving at a good enough workstation. This will also cover my thoughts on using cloud for AI/DL by individual practitioners.