yeah, i kinda lucked my way into my current position, all my knowledge that's useful for it came from undergrad QM. but if you know how large-dimensional linear spaces function and how to do things like change of bases, projection onto subspaces, etc. you can do most of ML, while largely avoiding a lot of the formalities of probability theory.
as for lucking into my way into my current position, we wanted to get into hyperspectral vision, and no one at the company was doing it, so as an intern i made some really crude python sklearn PLS implementations, that I guess worked well enough to start prototype testing. four years later I have an engineering title and i've moved my methods well beyond linear boundaries and have a ui program for training that I'm teaching people how to use now.
I want to present something along the lines of proposing that hyperspectral data should be thought of as an infinite dimensional hilbert space, but i work private sector and can't present my methods to anyone, including the rest of the company who don't really care what i do as long as it works. im probably already violating my nda posting this comment.
My professor said "welcome to Statistics 3", and he didn't lie
yeah, i kinda lucked my way into my current position, all my knowledge that's useful for it came from undergrad QM. but if you know how large-dimensional linear spaces function and how to do things like change of bases, projection onto subspaces, etc. you can do most of ML, while largely avoiding a lot of the formalities of probability theory.
as for lucking into my way into my current position, we wanted to get into hyperspectral vision, and no one at the company was doing it, so as an intern i made some really crude python sklearn PLS implementations, that I guess worked well enough to start prototype testing. four years later I have an engineering title and i've moved my methods well beyond linear boundaries and have a ui program for training that I'm teaching people how to use now.
I want to present something along the lines of proposing that hyperspectral data should be thought of as an infinite dimensional hilbert space, but i work private sector and can't present my methods to anyone, including the rest of the company who don't really care what i do as long as it works. im probably already violating my nda posting this comment.
I studied biotech, so I'm googling that dilbert thing you mention.
Great post though, and I agree with you that ML is just a huge meme within the private the sector.
(Would love to hear about your idea w.r.t to hyperspectral data and hilbert spaces however understand you can't share it)