Hastie and Tibshirani teach a free course based on this book on Stanford's OpenEdX (https://online.stanford.edu/courses/sohs-ystatslearning-stat...). I highly recommend taking this course or reading the book before delving into ESL. IMO, ESL is excellent as a reference, but trying to learn by reading it linearly is not an optimal time investment.
Now if only a similar course existed for Wasserman's "All of Statistics..."
There's a Youtube playlist[1] of recorded lecture videos by Wasserman from his CMU course that uses All of Statistics as a textbook.
I haven't watched more than a couple of mins of them (yet), so no idea how good they are (but the blackboard is quite hard to see in the recordings). However, it obviously doesn't have all the extra stuff that you would get in a proper MOOC.
If I’m not mistaken, Wasserman’s lectures are on YouTube under “Intermediate Statistics Larry Wasserman (CMU-36-705)“. You can find course notes and assignments for 36-700 and 36-705 on the web, which seem to use All of Statistics as the course textbook.
CMU has posted a lot of great statistics material beyond those two courses.
I'm a big fan of ISL - one of the best intro machine-learning oriented textbooks out there IMO. If you're looking for book that still offers a broad survey while going a bit deeper into the math, I recommend Elements of Statistical Learning as well (they share 2 authors):
Recommendation for ESL seconded. One of the best ML books in terms of writing, development of intuition, breadth of topics (ofc doesn't cover everything, esp deep learning).
What I love about the book is how the topics are "connected" so to speak. The narrative within a theme is typically "let's look at problem P, here's technique Q to solve P, but if you thought about P slightly differently you would see something like technique R would also work, so let's talk about that now".
The level of math might be tough for a beginner though.
I'm currently working through this book. Highly recommend, even if you have no intention on learning R. The R part is very limited, you will not learn R programming, but if you already know R, it is very useful to end each chapter with a practical demonstration of the theory.
I am trying to decide between going through a statistical learning vs. a deep learning textbook/course. Any thoughts on what would be more rewarding for someone with no immediate plans to work in ML nor do graduate level research. Thank you.
This is one of my all time favorite technical books. I wrote a review of sorts a few years back[0]. It doesn’t cover any deep learning topics, which perhaps dates it at this point, but it gives solid fundamentals on a breadth of techniques common in industry. This is always in my recommendation list for folks making the transition from more systems or product engineering to ML.
That said, if your choice is more general, statistical learning vs deep learning, I’m sure at this point you can find more approachable deep learning primers. This book just isn’t it IMHO.
I've been working my way through this book, and it's fantastic. I love the way this book grounds all the discussion of statistical learning with a practical data analysis problem.
I know R, but don't enjoy programming in it (although I love the documentation R has for its various libraries). Luckily there's various attempts to translate all the ISLR exercises into Python on github[1][2] that I have found immensely useful in understanding implementation.
Interesting. I like Python but don't enjoy programming in it. I find programming with R much more to my liking. I do data analysis and modeling mostly.
A useful practice, in my experience, is to implement R code samples in some other language, like C/D. Implementing lower level math functions for oneself can also be fun instead of relying on a library, depending on one's ultimate goals/interests.
I think that this is a great exercise to really learn the implementation details of the methods, but the point of ISLR is more to get scientists (and the like) up and running with the usage of these methods.
I'm a data scientist without a formal background in programming. Can someone please explain why implementing math functions in C/D is different than doing it in R?
For example, I would assume that creating a mean function using numbers and operators would be language-agnostic.
It's the difference between using a calculator and knowing how to multiply two 5 digit numbers with pen and paper.
You might say, "Oh but we can do that in R". Well, if you start doing that one function at a time, very soon you find yourself with R grinding to a halt.
I think Julia would be a better choice, especially since you can directly use R & Python from it. This way you can implement the methods in a scientific language yourself and still use the original R methods.
Now if only a similar course existed for Wasserman's "All of Statistics..."