How do I learn all the required math and statistics to be able to pursue machine learning career?
I already have the programming skills but I heard you need lots of math and stuff. Would pic rel be sufficient?
Thanks in advance.
How do I learn all the required math and statistics to be able to pursue machine learning career?
You need to start somewhere. You know how to program so you will figure out, after this book, if that's enough or not. You surely won't waste time reading it
this
stop asking the same moronic question every day you either do it or not
>Would pic rel be sufficient?
no. ML is rarely just theoretical. You need a good understanding of how of the common algorithms including their implementations. Data Science From Scratch is a good book for that. In addition, you'll need to know modern ways of serving these models.
That's the weirdest way to say "linear algebra."
>Reminder: bot is for discussing topics pertaining to science and mathematics, not for helping you with your homework or helping you figure out your career path.
>If you want advice regarding college/university or your career path, go to /adv/ - Advice.
this is more of a BOT literature advice thread
is linear algebra (whatever that is) enough for ML?
Legit question, why is there no proper statistics book for non-morons?
I'm not even talking about complete rigor or forcing measure theoretic-probability in statistics textbook.
Just something intellectually honest. Like limitations and coverage of the methods. Different ways to approach the same problem. Why every statistics books have completely different TOC. And ideally a lot of computer applications/exercises with R or python, considering the popular usage today.
It's like all statistics book are written for idiots (stats majors).
If someone mentions Casella Berger, I gonna kill someone.
A modern introduction to probability and statistics understanding why and how
Developing Thinking in Statistics
How to Confuse with Statistics or The Use and Misuse of Conditional Probabilities
Probability and statistics by example (Vol 1) - Basic probability and statistics
Advanced Statistics from an Elementary Point of View - Panik
I'm not sure what you're asking for. There aren't very many different ways of approaching the same problem. There are somewhat different algorithms, but every problem falls into the same bucket more or less.
If you want something comprehensive, I recommend Wasserman. If you want something covering basic programming, I recommend Introduction to Statistical Learning with R.
Anything else though literally requires entire courses to understand properly. You can (and should) spend an entire course worth of time just learning basic linear and general linear regressions for example (something that is typically given about a chapter in a standard statistics text and only a chapter in most machine learning texts despite being probably the most fundamental problem in statistics).
The reason everything is scattered and basic in statistics is because the field is frick huge. Literally every industry uses it and has different business cases and problems which leads to different interpretation and flows of methods. In industrial cases, for example, you're almost never working with a shit ton of data and EVERYTHING is geared primarily towards experimental designs for optimizations. It's much better to learn fields within statistics one at a time.
Also a lot of older texts, like casella and berger, don't properly integrate enough simulations.
>Introduction to Statistical Learning with R
How much practice does this book give you with implementation? Do you develop a solid grounding in R along the way, or would it be better to work through something separate just dedicated to R?
The authors recently released a Python version of that book, so you could read that.
Linear algebra and multivariate calculus is what you need
Not OP but I've already covered most math topics as I'm towards the end of my math degree. What kind of things to I look into now to lean towards the popular fields of ML and AI?
functional analysis
this book is too terse if you've never encountered the topics before. it's great, one of my favourites but only really when used as a refresher. start with geron's HOML book as a broad overview, and gil strang's 18.06 linear algebra lectures (and do the exercises from the book, i know you will want to skip them, do not)
bishop will give you a good foundation. you will realise that all of deep learning is just nonlinear curve fitting. with that perspective you can use anything to learn the practical side (i.e. HOML book). pretty quick you will be able to read any paper that interests you. congrats you can now fiddle around with model parameters, get a 1% increase in performance and publish your results at a top tier conference
>bishop will give you a good foundation. you will realise that all of deep learning is just nonlinear curve fitting. with that perspective you can use anything to learn the practical side (i.e. HOML book). pretty quick you will be able to read any paper that interests you. congrats you can now fiddle around with model parameters, get a 1% increase in performance and publish your results at a top tier conference
You must be grossly exaggerating, or average math PhD graduate excess that can't get jobs would overwhelm this field, like mostly physicists have done to biology.
it was definitely an exaggeration, more commenting that a lot of the research being done isn't that great and basically amounts to "we applied deep learning to [thing]"
Well that essentially is the field, sorry it's not math PhDs, but it's all people who came from industry. The field is primarily those who encountered real problems in data or optimizations and then tried to figure out the solution not so much the other way around. You don't really get goobers who have studied ML their entire life just to study ML so much.
Also, there are a LOT of mathematicians in ML world. It's a recognized field of applied math.
>import numpy as np
>The manifold hypothesis posits that all natural data lies on a low-dimensional manifold within the high-dimensional space where it is encoded. That’s a pretty strong statement about the structure of information in the universe. As far as we know, it’s accurate, and it’s the reason deep learning works. It’s true for MNIST digits, but also for human faces, tree morphology, the sounds of the human voice, and even natural language.
Deep learning is a good prompt for bad philosophy.
Math is a meme. If you want to make bank with ML practice and learn cleaning and organising data and the most popular libraries for building black box ensembles that get good results.
>how do i learn the math
ngmi
Get a math or physics degree. Physics would help you apply it, be scientific about models, and learn modeling as a whole
Or you could just study the math on your own and just be bad at it like most other people too
that book is pretty garbage for learning the maths to be honest. it's one of those books that says you should know all this maths, and then gives you a bad lesson in the maths, while telling you to use someone else's course to learn the maths. it's confused. if i knew that kind of maths already why the frick would i need it in your book? anyway.
herbert goss' course on mit ocw will teach you pretty much all the maths you need to know. linear algebra is tackled via vectors in the multi-variable section of the course and is very well done. actually the whole course is well done and structured for people to learn by themselves. it was intended as a refresher for people returning to do post-graduate studies but is very accessible, and the structured nature of it helps a lot. it's also important if learning by yourself that you have access to the answers to the material. that excludes quite a lot of books.
https://ocw.mit.edu/courses/res-18-006-calculus-revisited-single-variable-calculus-fall-2010/
vectors are used here as an intro to pure mathematics.
https://ocw.mit.edu/courses/res-18-007-calculus-revisited-multivariable-calculus-fall-2011/
https://ocw.mit.edu/courses/res-18-008-calculus-revisited-complex-variables-differential-equations-and-linear-algebra-fall-2011/
grinfeld i find to be a good lecturer but his materials aren't great for working from. i'd use him to help explain stuff you're having trouble with but get the bulk of your material elsewhere.