This was on HN quite some time ago:

http://pne.people.si.umich.edu/PDF/howtoread.pdf

I find it valuable for another reason: I'm always constrained for time, so I might not get through an entire book. If that book is deeply technical, it discourages me from continuing, because I have to take the time to re-skim what was read previously.

I somewhat recently started trying the above method. With multiple iterations through the book, I can get an idea of the information and at the very least know where it find it and that it exists. Depending on the iteration, I'll have some notes to fall back on, or have some important concepts underlined.

This is also why you shouldn't just watch beginner tutorials on complex software like 3d modeling packages or digital audio workstations.

Those are good, but you should also watch professionals use the tools on really complicated projects in person or in screencasts. While you won't be able to follow along fast enough and will be in way over your head, you will learn what's possible and be able to remember in the future when you run into a problem that needs complex radial symmetry, or whatever, that there a way that is much faster.

It also broadens your notion of what is possible in the application, rather than trying to reason from the ground up of the simple tools you have learned in a beginner tutorial.

That's a good guide, but one thing I think it doesn't emphasize enough is the preface. If a book has a preface, that's usually where the authors make their case for why the book exists in the first place. It's worth paying extra attention to the preface, because it establishes the context of the entire work. The ToC tells you what's in there, and it might hint at why, and the intro might give you more historical background, but the preface is almost always short and sweet.

How does this work for mathematic books?

Great PDF on how to read.

It is extremely inefficient to just read a book 5 times, it is like throwing mud to the wall hoping that it sticks. It's much more effective to focus on each topic and try to remember and repeat the informations just read. Having a general mental framework on the book on the other hand is very helpful.

Sorry, but you're wrong.

You're not reading a book five times, cover to cover. That's not a productive use of your time. Instead, you should breeze through the table of contents. Skim the first chapter. Still interested? Skip to the last chapter and read that. Jump around to the parts that interest you. Read through it again with a broad outline of what you're reading. Encounter a boring section? Skip ahead by twenty pages or drop the book altogether.

This is the best strategy for reading books for knowledge, and not just to say you've read a certain book.

I wouldn't say that reading a book multiple times is wrong. People learn in different ways. One way works for some and another way works for others. Anyway, finding the interesting parts of a book is a good tip, one of the axioms of learning is that you'll learn much faster if you are interested. Building interest on the book/subject by reading what we find interesting is an excellent way to learn. Of course you have to eventually read what you previously found uninteresting. Hopefully by then you'll have enough knowledge or curiosity to get thru it. What helps is to constantly ask questions as you go through the book and finding the answers will help. Generally speaking humans need to solve problems to survive. Our brain gets a hit of Dopamine every time we get an answer. So, by asking why as we read, we'll be rewarded by learning.

My tip is to make sure you understand the subject. By putting it in your own words. A problem with reading is that it's easy to fool ourselves into thinking that we know something when in reality we don't. We need to constantly test ourselves.

I agree that this scatter based approach to learning works very well for me, mainly because I am always interested in what I am learning. This approach also works for narrow scope based topics, by "flitting" between resources (internet links etc) rather than obsess about a single (possibly badly worded) article/tutorial.

I am reading "How to read a book" and the process you described is akin to inspectional reading. The book goes in great detail how to utilize the inspectional reading process and improve it.

It also describes the next step "analytical reading" which comprises of mostly asking questions about the topic in general and about the author's intent.

I'm sorry, but I don't think you have much experience in learning. In this way you will only remember a fraction compared to focusing step by step on the different topics and purposefully recall all the concepts, ideally putting them to practice starting from scratch without looking at all at the book. Source: pretty much every research about learning agrees that the conscious efforts of recalling, especially with spaced repetition is the best way to learn something. Plus anecdotally decades of experience in learning a huge amount of topics, programming languages, human languages and whatever. If you want to learn something useful I think that it may be helpful for you the course "learning how to learn", already cited by someone else here, in coursera that explains quite well why just skimming and reading a book is a very inefficient way of learning something. Accordingly to that course and a lot of the current learning research I'm not the one that is wrong.

Not it's not. It's not like each pass takes as long as if you were doing it in one pass. I would actually say that this approach is faster as it lets you focus better and on the things you want to get.

> it is like throwing mud to the wall hoping that it sticks

I must have not explained it well because that's the exact opposite of what I'm talking about. It's not about what sticks, it's about creating the connect-the-dots

> It's much more effective to focus on each topic and try to remember and repeat the informations just read.

Effective for what? Assuming that a neural net is a good approximation of the human learning model, why do nns tend to be be trained over multiple epochs?

To create the connections you have to consciously think and recall not only the topics, but also understand how they are connected and practice them. Seriously, I thought this was the last place on earth were I would have found people saying that to learn something the best way is to read a book 5 times. Because if you want to learn Java obviously the best way is to read a Java book 5 times, instead of reading one concept at a time, understanding it, recalling it, putting it in practice with exercises without looking at the book and connecting all the concepts. I will continue learning in my way that for you is for some reason so alien, good luck in learning reading 5 times.

You are misunderstanding what I'm saying.

I do a bit of a twist on this : I get as many books on the same subject as I can and read them all at least once. I typically find that each one has valuable nuggets, one really clicks with my brain, and all contribute to scaffolding a solid internal representation.

Haha same here. It's becoming a bit of an issue, as my apt is getting filled up.

> The first pass might take me like an hour

How do you fast read it? It's more like skimming it rather than actually understanding it. It would be frustrating for me, I'm more like a perfectionist who hates to leave behind not understood paragraphs. The idea is neat though, I lean the same way, but it's not true how I read books.

Very quickly. I would say I'm only scanning for terms that occur a lot and that I'm not familiar with. And in what order they are introduced.