literally saying, i am very interested in ML. I know just basic ML(masters university levels). but i feel like i need to be more mathematical person to be a first author of ml papers.(have one as second author in nips but i was just a supporter) instead going to phD, what can i do to improve my math skills? i know basic calculus but do not know high levels such as lipshitz, hilbert space etc... any recommendation to starts? feels overwhelming by following all that youtubes from the basic... thanks a lot folks!!
Linear algebra
I think i do know quiet enough for linear algebra. I can understand the paper that uses undergrad math level but cant create nor read theoretical one. I was very impressed by absteact of papers about multiplicative weight updates but could not understand a single page..
Lip shitz 😂 I’ll head back to leeting
Why do you want to first author a paper about a topic on which you are not familiar?
Wanting to contribute to a field is a good thing
Not if you’re not well versed in a subject. That’s how shitty papers get published that really don’t contribute much.
Find a topic/paper you really want to understand then go deeper as you come across things you don't understand. I did this with StyleGAN and by the end was able to read and understand the paper about 90%.
Slow down, adjust your expectations. Pick a paper, and make it your goal to get through it in, say, a month. Go line by line, looking up everything. Set an expectation that those YouTube videos you're watching will be where you'll spend your time. Eventually, you'll need them less and less, until you can read most papers without encountering too many unfamiliar words.
Eh if you know ML at a masters degree level you’re probably fine; oftentimes the math in those papers is added after the fact to make the paper look more substantial
Post a paper you're interested in and tell us where you're getting stuck. I'm assuming you're targeting more theoretical stuff like in COLT
You don't need to be a "mathematical person" to get first author on a paper in a top tier ML conference. Just learn the math as needed for your specific area.
Real analysis. Sampling theory. Vector calculus
Any recommendation to follow those three topics? Thanks by the way!..is reading textbook or watchong youtube is all i can do?
Rudin, Schervish, Spivak