The Mathematics of Machine Learning 

*Picks up Deep Learning by Goodfellow, Bengio and Courville. Doesn't understand a word. Orders Calculus for Dummies.*

andrew ng

MachinLearning 2020! IN MY CHANNEL

I really don't like how courses and videos just say it's ok if you don't know calculus, statistic, linear algebra... Dude, it's not a monster, everyone can learn it. So go back, study the pre-requesites, and then you will actually understand how machine learning algorithms work.

🤣🤣🤣

Susan Wojcicki is showing her true colors as an unintellectual SJW by her choice of what she demonetizes.

Now TikTok is better than youtube

for real, zach don't fuck around lol

MachinLearning 2020! IN MY CHANNEL

@@landosllim4576 Zach ain't bitch made.

Thanks man! Really appreciate the comment and sharing the video, helps a lot.

@ Cameron McWilliams ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Cr6VqTRO1v0.html

MachinLearning 2020! IN MY CHANNEL

This has been one of your best videos. Is the math you are using with the examples of the price of a house and passing or failing a test statistics, calculus, mathematical modelling, or probability?

Can all the questions you used the computer to help you answer be answered by doing the math with a pencil, a paper, and with or without a calculator? Do you have to know graph theory to be able to solve and or answer problems like the examples you showed in this video?

Hey major prep which video editing app do you use??

Premier pro

@@zachstar hey could you please tell which application do you use to create these videos

Halo 1player?? You got discord?

Any good PDFs out there?

Jonathan Molina Linear Algebra and Its Applications...search for this book there is a free pdf ...

I couldnt find the download :/ is there a link ?

Agree

The class he mentioned, Andrew Ng's Intro to Machine Learning, is an incredibly well taught machine learning course with videos, quizzes, programming assignments, and a certificate. I highly recommend you take it.

@@camozot thank you so much for the info!!

This video is sponsored by Coursera dummy guess where you will find the course for this

I'm thinking too

MachinLearning 2020! IN MY CHANNEL

UPDATE: Machine Learning and my other big CS interest, parallel programming, have pushed me to learning Linear Algebra SO much that I made that the first advanced math class I'll take after getting past Calc 2. (I can choose between Calc 3/Multivariable Calc vs. Linear Algebra vs. Differential Equations)

@@projectjt3149 I thought it's recommended you learn both Linear Algebra and Multivariable? I've already taken the former along with probability, and I have to take calc 3 this summer before ML in the fall

@mathsonboard It does. It introduces a little bit of algebra and the course walks you through a lot of projects on different kinds of machine learning, ie computer vision. Unfortunately, the course is taught in Octave, but someone did make a series of Medium articles where they wrote the Python version of the code.

lol I was thinking the same thing, he literally gave the exact explanation I so needed to finally understand this shit.

Machine learning ur needs soon they will tell you what to do

Go to Coursera do do the course for free Go to the stanford's one, that were he took the examples and the graphs from.

RU-vid algorithm machine learning

Btw weird question... Since you have a 3.8 gpa are you looking to study in ivy league schools, for your masters degree?

Thanks for the comment! And I probably wouldn't put priority on going to an ivy league. Maybe I'd apply but I think UCLA would be a top choice for me because it's a good school and I live close so would be a little more convenient.

If you're losing hope as a hs student just wait until college lol.

I second this

theoretical :) Happy to help.

Computational physics sounds like your kind of thing.

Unless you plan to do a PhD in somewhere like Switzerland or The Us don't Even think of experimental

Check out domain of science

At my university the prerequisites of machine learning course are linear Algebra, calculus 2 and probability which are all covered in EE program :) But if you are still a student in sophomore year I suggest you to switch to computer engineering since studying EE will leave you have little to no time for other interests such as ML unless you are great at managing your time.

Based on what I've seen, yes the math I learned in EE was plenty to get started with machine learning since I took up to multivariable calculus, linear algebra, and a statistics class for engineers. Later on there's probably more theory especially in statistics but the math that most engineers see is plenty to get you started.

@@zachstar Thank you very much for your reply. That's really good news. I love how engineering has so many transferable skills.

@ S M As long as you know linear algebra, multivariable calculus, and statistics that is a good solid foundation to start. Knowing some probability theory would be beneficial as well.

It also depends on the field you are actually interested in. For example, when I was taking digital signal processing courses I had an intense introduction to stochastic gradient decent optimisation, neural networks, and a lot of statistics. So if you want to be prepared to proceed with ML in the future, DSP courses will be very useful.

You have seen nothing not even a single.drop of maths my boy..

Awesome to hear! Keep that passion and drive, it will take you far. You'll see people like Sheela who will try to hold you back. Don't let em.

Start with linear algebra and then matrix manipulation

I am an idiot. But power to you! Just use your powers for good, when you get there. Math can be fun, with a bit of brain conditioning. I just learned that too late in life. Good luck.

Good Passion my boy

Yup, that's calc1 right there

Yeah good question cause I didn’t go into too much detail on that. The error function was the sum of (expected output - predicted output)^2 (its a bit more complicated but this is the bulk of it). You take the derivative of that as normal, by moving the exponent down and subtracting 1 from it then apply the chain rule. Since you are just taking the set of points you can use a matrix to do this all. But you can treat the error function like an actual quadratic function essentially.

It must be some quadratic because the error value (for a given slope) is (a sum of) squares (of the difference between the actual value and the projected). That is a mouthful but nothing in that mouth is of higher order than quadratic. The thing I don't see yet is how to get the parameters for this quadratic, especially from the first value--as I think is suggested.