Ask HN: Resources on Modern ML Modeling Principles/Philosophy/Concepts?
I’ve been tasked with an objective to develop a system to report on user efficiency proactively. More specifically, use often overlooked data points to reveal additional information about business metrics and other deliverables. The goal being that this system is used as a global reporting tool, with all other existing reporting tools (reporting to this system and ultimately) becoming supplemental.
As this is a brand new world to me, I’ve been hitting the proverbial books (web articles, webinars, online courses, etc) pretty heavily and have formed a lose concept of the tools that might be used when and where, and have identified a number of data points relevant to this use case.
During my research I’ve noticed a difference in availability and use of tools and practices over the years. Many of what I’m finding is publications or courses on a particular platform or workflow of this sort of endeavor, so it’s very much expected to some extent.
What I’m having difficulty finding is the philosophical/conceptual stuff.
Rather than what is most popular or most cutting edge, Im looking for the “why” or “how” behind it: - The unforeseen obstacles that others have faced - what caused the previously considered best practice to become sub-par. - Emerging unpopular opinions in this space, and their reasoning, etc. - reasoning behind changes in what is considered best practices / industry standard
In short I’m looking for some information that I can use to build a set of principals to follow while working on this objective, and by future resources assigned to this objective.
Please assist!