This valuable review article on machine learning is almost a decade old but is still an insightful read even today.

Pedro Domingo eloquently authored this masterpiece on important lessons in machine learning a few years prior to the popular book, The Master Algorithm. The invaluable 12 key lessons of machine learning not often discussed in academic works or textbooks include only a few that seem outdated or mundane (“learn many models, not just one”, “intuition fails in high dimensions”, and “overfitting has many faces”), but concomitantly include some very salient ones: “it’s the generalization that counts”, “intuition fails in high dimensions”, “more data beats a cleverer algorithm”, and my personal favorite, “feature engineering is the key”.

This latter lesson stresses the importance of domain knowledge and expertise so often underestimated (especially in the era of deep learning and automated feature engineering), and yet this human-machine synergy is so critical in collaborative efforts in healthcare data science projects. He even provides a word of caution with automation of feature engineering.

Overall, this entire article is extremely readable even for the data science novice (just like his book) and is worth reading once in a while just to remind ourselves of the major tenets in machine learning from one of its master gurus.

Read the full masterpiece here