Fitting your first Logistic regression or Recurrent Neural Network (RNN) based model on a well-behaved, curated dataset will feel like magic until you are faced with a complex real-world business problem with a really messed up dataset or even worse, no dataset at all. Most ML practitioners, especially the ones from programming backgrounds, wonder why they hit a wall in their ML journey. This is entirely on the current ML pedagogy in circulation, a method-centric or technique-centric one. A technique-centric pedagogy to applied ML strips away the mandatory rigour to deliver on some narrow promises. It took me 108 submissions to come in the top 1% on my first Kaggle playground competition. This is what happens when we don’t have a good hang of the principles. So please take my word on this. 

“As to methods, there may be a million and then some, but principles are few. The man who grasps principles can successfully select his own methods. The man who tries methods, ignoring principles, is sure to have trouble.” — Harrington Emerson

Paraphrasing Harrington, if you know the principles you can tailor problem-specific methods to creatively apply ML. In hyper-derived fields like ML, getting a good hang of principles is your singular competitive advantage over your peers. Here are some compelling reasons:

Compelling case for a principle-centric approach

Upshots of a Principle-centric approach


A principle centric approach offers a solid transition path into a seasoned ML practitioner (or even an ML researcher, based on your career interests). ML practitioners are the ones who apply ML to solve real-world problems, whereas ML researchers are the boundary pushers looking at the next frontier of ML itself. GAN's fame Ian Goodfellow is a boundary pusher. But contrary to conventional wisdom i.e. Instead of thinking ML Practice and ML Research are two separate career paths, thinking ML Research as a natural progression to ML Practice has a better utility. A principle-centric pedagogy can help you move through this spectrum with ease.

Entry-level ML Practitioner → Experienced ML Practitioner → Seasoned ML Practitioner → ML Researcher