[eBook] The Elements of Statistical Learning 2E and Introduction to Statistical Learning (Currently Free)

Moved to Forum: Original Link

I have noticed there have been a lot of freebie stats/machine learning amazon books and udemy courses. Seems like many people are interested in these topics so I thought its worth mentioning two books which I have found very useful (and so have many others) and are available for free. These books are also highly regarded in the field.

Introduction to statistical learnings is a great book for people to learn about different techniques in machine learning. It covers how to apply different machine learning tools, as well as providing some theory about how they work. It requires basic knowledge of probability and maths (high school level). It also introduces you to the R-programming language which is widely used for statistics and machine learnings (especially in health and biology).

Elements of Statistical Learnings provides much more detail about various statistical/machine learning tools. It requires more maths knowledge (~1st year uni) to understand how many of these tools have been developed.

If you are more interested in epidemiology and clinical trials related statistics this is probably not the best book for you. There are other better resources on these topics.

Interestingly the Amazon kindle editions for these books cost $50+ and $70+. I suspect they will remain free for the foreseeable future on the author website but you never know…

Enjoy
DS

Related Stores

usc.edu
usc.edu

Comments

    • +4

      Then I would have missed it…but now you posted your comment, thank you for the link to the forums area I was unfamiliar with.

      • +1

        True that. Had OP not posted this here, I would have missed it too

  • Thanks!

  • Awesome! Thank you OP

  • Thanks for sharing OP.

  • +7

    These have always been free. Not sure what high school and uni OP when to but the maths in ISLR is at least 2nd year stats, and some of the proofs in ESL require at least 3rd year stats/econometrics…

    • +3

      Yep. Fair points

      I have basically had no formal education in statistics or maths after high school. I did health sciences undergrad degree, where we did a bit of clin epi/stats related to clinical research but that's it.
      I went onto do basic science/biology research and ended up working with lot of data, where statistics and machine learning methods are used frequently. I found ISLR very helpful to get better understanding of the theory. There are some bits where it gets bit technical from a maths perspective but the author does flag these as being "optional" and you get still get good understanding of the intuition behind what's going on without fully understanding the maths.

      I must concede that ESL is tricky and I do find some parts challenging to comprehend. I use it more as a reference book.

      I think one of the good things about these books especially ISLR is that it doesn't oversimplify what you are doing.

  • +1

    any free statitics books that can complete this set?

    • +4
      • Is it just my eyes or this book uses grey fonts?

      • Thanks, though I'll probably get introductory materials before reading this one.

      • Yeah, this is a really good one!

  • the best textbooks!

  • +4

    These are great books for ML. If anyone is looking for good maths book for ML, this one is good one https://mml-book.github.io/book/mml-book.pdf

    • Heard good things about that book.

  • +2

    These books have been freely available since their inception.

    Here is my review of them (Currently doing my Masters in Statistics).

    "Introduction to statistical learning" is a decent book if you've done some first year uni probability/stats course. It is very much an overview of popular methods in statistical learning over the last decade but it doesn't give you enough knowledge to go out and start using the material on your own. It is a good starting point nonetheless.

    "Elements of Statistical Learning" used to be a great textbook but now it is too old and in dire need of an update. Last edition was made in 2009 and by current pace of statistical learning it may as well be a Dinosaur (the section on Neural networks is extremely outdated). Just given how old it is it is outclassed by a lot of newer books published in the last 5ish years. I removed it from my bookshelf.

    • +1

      What do you recommend?

      I don't really do much with neural networks/deep learning but for most of the methods used in processing large biological datasets, these books are really comprehensive.

      • +1

        For EoSTL? I'm not sure about free, but "Data Science and Machine Learning : Mathematical and Statistical Methods" is an example of a more modern reference text—not only is it rigorous, it also provides algorithms and code in python. It has sections on Monte Carlo methods, and provides Neural Networks in the context of Deep Learning.

        As a bonus, it was published by professors at Uni of Queensland.

        • Thank you for suggesting this book, I am getting it now. Hopefully, it will cover most of the math/stat aspects of Data Science.

    • too old and in dire need of an update […] I removed it from my bookshelf.

      Many would argue that you made a mistake.

      This is one of the best and highest cited theoretical books out there. There is no such thing as "Dinosaur" for the theories the book describes, and even the newest ML frameworks rely on these very principles.

      Furthermore, in 2020 there are already over 800 citations to this book, which would be pretty hard to explain away if you were correct.

      On the other hand, your recommended "Data Science and Machine Learning: Mathematical and Statistical Methods" book has practically no recognition in scientific research

      • On citations: Alpaydin's "Introduction To Machine Learning" covers the same area as EoSTL. That book has more citations (total) and almost 700 citations this year too. It was by far one of the worst books I've ever read (but at the very least he bothers to update it every few years).

        Nothing in EoSTL is outright wrong it is just dated. It is missing a lot of things that we've learned about (or has exploded in popularity) over the last ten years. Example: double descent in Bias-Variance trade offs, t-sne, Monte Carlo methods, Deep Learning, Bayesian Optimisation etc. EoSTL is still a great book for the material it covers but the fact that its last edition was made in 2009 and that means you are going to get an education from 2009. Statistical Learning has changed significantly over the last 10 years.

  • Since you mentioned it in your background OP, do you have any recommendations for books to gain a basic understanding of statistics in clinical trials?

    • +1

      This is one place I’d start.
      https://peterattiamd.com/ns001/

      Peter Attia gives a pretty good rundown for someone with no assumed stats background.

      • Thank you! Will check it out.

        • +1

          Although Peter gave one of the wrong definitions of the p-value, so read this with caution - and perhaps reach out to statisticians who can help you out with your specific questions :)

          The p-value is trying to answer the question: What is the probability […] of rejecting the null hypothesis when it is, in fact, true?

          This is not the definition of the p-value

          If not, the results are deemed not statistically significant, and the null hypothesis is accepted.

          And this is also incorrect - you never "accept" the null hypothesis, merely say that in this instance, you did not find evidence against it. Sigh.

          Yeah, just reach out to statisticians with your questions

          • @ocoolio: Thank you for the clarification. I think I might have to do some reading the appreciate the nuances here.

  • +1

    This has been free like forever

Login or Join to leave a comment