In school, I hated math and thought it’s very useless for everyday living (which is probably wrong, but I definitely never missed math so far). The end result is, that I don’t know any math besides additions, subtractions, divisions, and multiplications.
I’m wondering what math knowledge I ideally have to study the course.
I’m planning to use Khan Academy’s math courses (if nobody has a better suggestion). It would be great, if someone could help me select the right courses!
If I need all the math that is taught in high school: Up to which Khan Academy course would that be? (I’m not from the US, btw., so I have no clue what is taught in the US).
I searched on the forum and on Google for what the exact math requirements are, but I couldn’t find anything.
So this information might also be useful to other students.
I think if you’re using places like khan then they normally start by explaining the prerequisites you need to understand that video so you can just work backwards if necessary.
Do you know (roughly), when I need these individual math skills? I’d like to learn them “Just-in-time”. The first lesson had no math, for example.
How long will it take to learn these topics enough? I know, it’s kind of silly to ask. But maybe you can give me a very rough estimate (i.e. if it will take days/weeks/months).
I’m not 100% sure if this is relevant, but as a second data point I can offer that I’m programming for 15 years (mostly high-level languages like Python, JavaScript and others). For the last 8 months I’m using Rust (which is not the easiest to learn language…).
Not sure exactly when as I did last year’s course which is more or less but not exactly the same.
How long depends on you, what you already know and how quickly you learn, so it’s impossible to guess.
Like I said, go and get a basic understanding very quickly and dive deeper if and when you feel the need. If you only want to know how to use deep learning without a full understanding of how it does what it does then you might not need to go as deep with the maths as someone else so I wouldn’t worry too much at this point.
I think the first thing I would say is don’t panic! I was in a pretty similar place last year and managed to get through so definitely don’t feel that you can’t.
I think, at least for part 1, the most important things are linear algebra and calculus.
(Specifically: matrix multiplication especially the shape of the output, and the derivative of somthing)
Khan Academy is a great resource but I would say that having an understanding/ intuition of the maths is way more important than knowing how to solve something on the page.
I think the best place to start is to have a look at 3blue1brown. He has a couple of great playlists which helped me get an understanding of the maths behind things.
I think it is in part 1 but http://matrixmultiplication.xyz/ is great for understanding what matrix multiplication actually does.
Besides that I would say get started - the best way to find out what you need to know is to start doing things and finding out the hard way! Even if you don’t 100% get something, still try and do it - it’s really tempting to not start something and instead do more background learning to make it easier. I’ve definitely done this! But in my experience it’s both quicker and better just to give it a go and learn by doing.
Apart from that, I would only say don’t get disheartened if you don’t get something the first couple of times (I definitely had to rewatch plenty of things I didn’t get!)
If the above has too many things to suggest, why don’t you try to do the image classifier first? with it, you don’t need any math at all! Fastai is about play and do first and learn.
@lukemshepherd: Wow, thank you so much – these playlists look fantastic!
Linear algebra is more interesting than I’ve expected. I’m interested in programmatically generating vector graphics – and linear algebra seems to be the fundamental basis of this.
(Sorry it’s a bit long!)
Not at all too long! Everything was helpful and seems important!
@JonathanSum: Thank you, I’ll keep that in mind. I think I will go through the above playlists (so far they seem unexpectedly interesting to me), and then through the fast.ai course. If I don’t understand something (and this can’t be remedied with supplemental research on-the-fly), I’ll write the topic down and study it in detail later (after which I’ll repeat the lesson).