News for hackers. Stuff that matters.

Ask HN: I want to teach myself Machine Learning. Where should I start?

I just graduated from high school this past June, and due to a number of factors, I have elected to take a gap year and not attend college yet. I'm technologically fluent, but I don't have a lot of programming experience (I plan on learning more as I go, though.)

I want to learn to create, experiment with, modify, design, train, etc-- neural networks and machine learning models, but I don't know where to start. Rather than just blindly trusting Google search, I thought I'd ask HN since many of you do this stuff for a living!

Machine learning is so, so exciting to me, and and the possibilities with it seem endless. What resources would you recommend I dive into to teach myself about it?

I'm being vague here because I know there's a lot of different branches in the field and I'm still not sure which ones I'm most interested in.

That said, image/audio/text manipulation or synthesis and simulation are of particular interest to me (I'm a big fan of Two Minute Papers)

If you took the time to read this, thank you! HN has changed my life and introduced me to so many new ideas and resources already :)

Looking forward to any and all responses.

EDIT: Realized it might be worth it to mention that the current extent of my knowledge on the topics is limited to "grab a big dataset, create a model, refine the model through some kind of regression, then test with unfamiliar data." I also understand that there's a tooonnn more than just that available (i.e. GANs/synthesis etc).

( 8 comments | 13 points | on HN )
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by st1x7 on Thursday December 10, 2020 @04:13AM (#25371547)
I know that you've decided to take a gap year but my advice for next steps is actually to take undergraduate courses in single variable calculus, multivariable calculus, linear algebra and probability theory. All four of these are available online for free (course names: MIT 18.01, 18.02, 18.06, Harvard Stat110).
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by grassgreener on Thursday December 10, 2020 @12:24PM (#25375976)
This is the most important. If you really want to get into AI and be able to deliver insights, there is no short cut, you need to be fluent in Maths and Statistics.

This will open you many doors too, who knows, maybe in 3 years you will have decided that ML is not something that interest you anymore. Having a strong quantitative background is a very transferrable skill.

It's going to be hard and sometime very abstract, but it's worth it.

Please don't waste your time on tutorials.

http://nirvacana.com/thoughts/2013/07/08/becoming-a-data-sci...

Re:Ask HN: I want to teach myself Machine Learning. Where sh... by mkaic on Thursday December 10, 2020 @11:17AM (#25375265)
Thanks for the recommendation!
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by giantg2 on Thursday December 10, 2020 @09:01AM (#25373749)
Coursera has a free AI for everyone course by Stanford/Andrew Ng. I think they have some more in depth ones that are free but you wont earn the completion certificate.
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by mkaic on Thursday December 10, 2020 @11:16AM (#25375251)
I think I've heard of this. I'll definitely check it out, thanks!
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by ariosto on Friday December 11, 2020 @05:28PM (#25392290)
Any love for https://www.fast.ai/?
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by rumpelstiel on Thursday December 10, 2020 @03:51AM (#25371378)
i would say, get into Computer Science as soon as possible. The basics you learn there are essential, and then doing the fun stuff gave me lot of aha's
Re:Ask HN: I want to teach myself Machine Learning. Where sh... by rumpelstiel on Thursday December 10, 2020 @03:54AM (#25371398)
its like having lot of tools and even know the pieces of it, ... later you use them intuitively, that is way more fun, but might also be hard at the beginning. (math)