>>12119909>>12119863I'm shit at math and basically know nothing about any math field, but I think I have enough of an understanding of ML to apply it to some problems to a reasonable degree. I'm obviously never going to do research in it (maybe unless I study math hardcore every day for the next 10 years), but it's a little like CS and software development. You can really get by on the basics. Learn the few key pieces, and you'll be able to achieve your goals, probably.
So I'm skeptical that you need a PhD or even the breadth and depth of knowledge you recommend. Every additional resource and concept surely helps, maybe even by a lot in some cases, but if we're talking about efficient use of one's time, I don't think it's a necessity if you just want to be the person picking architectures and algorithms instead of inventing them.
I say this because it's the applied side that'll really require most of your time and effort, if this is for a job or for-profi thing you're making. Doesn't matter how good you are at the rest if you can't both apply it and, more importantly, know when to apply it and what problems to tackle in the first place which may have some aspects that are amenable to it. There's too much to learn in this world to spend a year just studying theoretical shit unless you plan to specifically work in a theoretical capacity or unless you're a naturally very fast learner and can effectively juggle learning both sides at once.
