Numerical Linear Algebra for Programmers

That looks great, thank you. I’m actually tempted to buy it - have you worked through it?

This looks perfect for my case. I think I"m going to order this, thanks!

I watched it, still either too abstract or too much doing math for the sake of math, not enough explaining why I would want to take the inverse of a determinants eigenvalue or whatever the fuck linear algebra is supposed to be about, because halfway through the course I still had barely a clue what the point was.

I can see the appeal of the lower barrier to entry, but I would need to follow the project every month to see if they’re delivering. Really don’t want to be burdened with that.

Yeah, well, that's a nonsense comparison: the neanderthal package is freaking lapack (in C++), and they used a java trick to get Clojure fast on it (basically run the C++ on allocated unmanaged memory blobs, which is the only marginally sane thing one can do in a language like Clojure). I wrote an early version of this in Clojure before throwing my hands up at the atrocity of doing this sort of thing. They also wash over very obvious (to anyone who knows how a computer works) issues with using Cuda and Clojure numerics in general. Misleading and Fail statements.

If the whole book is filled with such chirpy and wrong assertions, they should fall on their swords for offending their numeric linear algebra forefathers and the shame of it all.

> I don't like Strang books that much as I find them a bit verbose

I'm not sure I agree with you. I remember his elementary calculus book, where he states and proves the fundamental theorem of calculus on the first two pages (including the definition of derivative and the definition of integral). He says something like "this is not a car analogy, this is the real thing and we are already done, the rest of the book is just minor details and examples". I do not know how you can be more concise than this, considering that half of the first page was taken by a cute drawing of a speedometer and an odometer. That you call verbose?

Surely the issues of "publisher as middleman" and "book-by-subscription" are orthogonal?

Best books are often written by academics, who get paid to work on them by their university. The royalties are just some nice extra income.

Exactly. Books have always been subscription based. Difference is it’s not monthly. But for that monthly subscription you get so much. Plus the good feels knowing you’re supporting some great open source work.

I've bought plenty of MEAP books at Manning that were incomplete. The difference is that I only had to pay the $30-40 (normal book price) one time and never think about it again, and new revisions would just be delivered to me.

There have been several early access books where the author needed to take a (well deserved) break, or be slow at delivering new chapters because of other revisionary work. This has sometimes meant a few months without new content, which I was OK with in that model.

In this model I could easily end up spending way more than than that, and every month I have to make a decision if it's worth cancelling.

Not only that you have another book on there that's "early access too" so you are juggling two books (and books are not trivial things to write and get finished/polished). So how am I to expect you'll get either done in a reasonable time.

So sorry, I'd be happy to give you a flat $30 for early access to the book and full access to the ebook when it's finished, but I'm not going to subscribe and hope it's in a good state prior to reaching that point in monthly fee.

This is despite I think this book is very much what I'm looking for as I'm wanting to get into 3d graphics, but I'll pass.

weird game theory view here, would you not be incentivized to never finish the book because people might be waiting for that 1.0 moment to cancel their subscription?

You should explain it clearly on your page, not here.

I also was confused by the model. The model you try to use is obvious for you but in order for people to understand you need to make an effort in communication, because most people are not used to it.

You have to be extra careful because people have fear of the unknown, so you need to also get rid of people's fears, you do it here("if you don't like it,just don't continue the subscription"), but you should do where people take the decision.

I think the model makes sense but you need to explain it in the buying page as a benefit they we will be getting. For example:

With this subscription you will be supporting the development of free open source libraries that will become yours forever.

Now I see that you display this information but I did not see it while being interested on the book, so it is not visible enough.

If I were you I would not try to train my customers, it is too much work, I would offer the book plus code plus 8 months of subscription for just $119!!

Then in the subscription I would give my customers something they can not get easily out of it, like the network connections and support with people interested in this area all around the world(the people that buy your book), so once they are subscribed, and they are used to it, they have to remain subscribed.

The value for me of those open source libraries is zero because I don't use them. Only if people use them, they can value it. You need to get people in the loop first.

Do you plan on just selling the book in the traditional way once it is done? I really have no interest in books-as-a-service.

If a significant part of the value is the libraries, then you must mention that more prominently. Is the goal to write/sell a book with code examples, or to sell a book subsidizing the development of open source libraries?

Good comment, my initial thought after reading the parent was that it was by the same person.

Clarification: the alpha book you're referring to is not connected to me nor this book.

> The reference you are looking for (includes executable C++) is Numerical Recipes by William Press et al.

I bought this book hoping it would be to floating point what Hacker's Delight is to integers.

It is not that.

I'm not really sure what it is. Most of the numerical algorithms I wanted weren't included at all. Those that were, like random number generation, were out of date and covered with much less rigor than you can find online with some Googling. IIRC their algorithm for generating a random float in the range [0,1] is just wrong.

The book does indeed include executable C++. But the C++ it includes is a travesty. It's C++ written by someone who sort of understood C.

Anyway I did not like it. :)

Trefethen&Bau, Golub & Van Loan and Demmel's "Applied Numerical Linear Algebra" make up the holy trinity. I don't know what audience the OP book is supposed to serve, but it's evident on inspection; it ain't me.

Interesting. Why not just use C?

I can understand why C++ isn't appropriate for this. Object-oriented programming has little to do with linear algebra, and a nightmarishly-complex language that's desperately trying to reinvent itself also isn't appropriate for pedagogy. But you still need a lingua franca, a language you can use to express the relevant principles without trying to evangelize one arcane functional language or another at the same time like so many authors try to do these days.

Being a relatively straightforward procedural language, C works well for this purpose. Everybody doing numerical computing is at least familiar with the fundamentals of C programming, or should be.

Is there some institutional resistance to plain old C, or is it just a matter of widespread personal prejudice among both students and teachers?