0

I would like to buy a book about AI and neural networks written on accessible level for a 17 years old mathematically very gifted student interested in these topics. The book should contain some sections about perceptrons and optical character recognition. I am aware of https://www.deeplearningbook.org/ but it does not fully satisfy me. Mostly, because it goes too slow and too long. For instance, in order to grasp the back-propagation algorithm one needs to read 60 pages! Eqs.(6.49-6.52) are particular shocking, I thought every student should know rules of chain differentiation! We do not write such trivial things in, e.g., theoretical physics.

Now, Internet is full of all possible blogs and tutorials, but it is impossible to filter some nice and concise exposition for people with mathematical background. I notice some popular extremes such as i) prolonged discussion of a single neuron, ii) XOR example, iii) very technical tutorials, which require a lot of python, a lot of packages, web-servers etc.

I am seeking some nice text to create a neural network completely from scratch, and with some impressive performance for e.g. written digits recognition. No reliance on external packages, no object-oriented features, as it may deter young students. But functional programming paradigm is welcome.

Therefore, I was thinking about more refined and concise books, preferably with good typography and illustrations, hard cover, suitable as a gift for a mathematically-inclined student. What would be your recommendation?

yarchik
  • 67
  • 3
  • 1
    Please, first rewrite the title in the form of a **specific question**. A specific question is something like "Do you know of a book that satisfies requirements X?". Second, please, go through this https://ai.stackexchange.com/q/3647/2444, and let us know if it satisfies your requirements. – nbro Jul 07 '22 at 15:46
  • 1
    I didn't downvote. Second, you should try to be more polite when people try to help. Third, your question should be specific but not lead to opinions. Currently, the title says "a beautiful...", beautiful is subjective. So, again, try to ask a question that will to faccts. – nbro Jul 08 '22 at 21:25
  • 1
    While this is a moderately interesting question, I don't believe it fits here in this stack. – David Hoelzer Jul 09 '22 at 21:56
  • 1
    @DavidHoelzer Why do you think so? Mathoverflow with >130k questions is full of questions about sources of theorems and books. Stackoverflow with >22m questions contains alone 288 posts about "Book Guide" recommendations. The `The Definitive C++ Book Guide and List` was upvoted 4234 times! In cs.stackexchange with 44k questions the question "books for algorithms" was upvoted 23 times. Yet, on ai.stackexchange with just 10k posts, questions about book are not tolerated and frown upon. I do not see the logic. Is it that this field is so rapidly developing that no one has time to write books? – yarchik Jul 12 '22 at 08:10

2 Answers2

1

My vote would go to Artificial Intelligence: A Modern Approach. It is not concise, but that's a feature, just select the chapters you are interested into.

Rexcirus
  • 1,131
  • 7
  • 19
0

After a lot of searching the following seems to be a good choice, but sometimes with repetitions of material. The math is very accessible.

Neural Networks and Deep Learning: A Textbook by Charu C. Aggarwal

yarchik
  • 67
  • 3