# Books that contains exclusively math problems/assignments in Deep Learning & Neural Networks

I am doing a Deep Learning Course.Suggest some books that contains exclusively math problems/assignments in Deep Learning & Neural Networks. I can understand that majority of the replies suggest "Deep learning", by Goodfellow, Bengio and Courville, which focuses on math, mostly linear algebra and calculus. But I am looking for books/articles/"assignments from universities" that contains/deals with EXERCISE MATH PROBLEMS in DL & ML. For example I have been given the following problem which I have uploaded (i dont except the solution, I'll find it somehow):

Since I am new to DL & NN, I want to know HOW & WHERE to go about, understand the problem & find a solution for it. I am not interested in searching AI chatGPTs for the answer.

From your problem example it seems an entry level neural network course to find a decision boundary of a single perceptron for linearly separable training data. Hagon's book Neural Network Design (2nd edition) specifically addresses this problem on page 4-5 and it also has many solved problems.

The decision boundary is determined by the input vectors for which the net input is zero

So for your case the decision boundary is $$w_1x_1+w_2x_2+bias=0$$

And on page 4-7 it has an intuitive diagram of the decision boundary for AND function, you OR function is very similar. So intuitively you can immediately see the only possible correct answer is c since $$x_2=-(w_1/w_2)x_1-bias/w_2$$ which has the right slope and $$x_2$$-intercept. Later the book also dives into the supervised training rule for single and multiple perceptrons which can be generalized later.

From your comments it seems like your course is covering topics related to feature descriptors in the context of image or signal processing. These are fundamental concepts in computer vision and signal processing which is not what this site focuses on and you may try other closely related sites.

In computer vision, a feature descriptor is a representation of an image region or object that is used for tasks such as object recognition, matching, and tracking. In this context, a signature usually refers to a representation or descriptor that characterizes a specific aspect of an object or pattern within an image or signal. In image processing, a co-occurrence matrix is a matrix that describes the joint occurrence of pixel intensity values at different spatial offsets in an image which is often used for texture analysis.

• I have found your reply much helpful. I hope this forum will help me find answers to my questions. Thank You
– GKK
Feb 23 at 22:30
• @kreetykishore Thanks for your positive comment, if an answer is helpful to you I encourage you upvote it or even accept it if satisfied :) Feb 23 at 22:49
• i tried to upvote, i have only 11, it is asking for atleast 15. how to accept?
– GKK
Feb 24 at 5:56
• Oh, gosh. I believe there's a check like icon below each answer point to accept which as original poster should have right to check. Feb 24 at 6:14

In that case I would recommend to look for courses offered by universities and solve their problem sheets. For example, have a look at this course Neural Network Theory at ETH Zurich. You can also find past exams if you want to solve them by your own, you can find them here. Also I can recommend The Science of Deep Learning by Iddo Drori.

But if you are specifically looking for "school books" the best way to go is to look for content of Deep Learning courses offered by universities. You will likely find literature that supports these courses. Also, have a look into MIT's Theoretical Foundations for Deep Learning course course.

• Thank You for your reply. I hope I can learn something from the sources given by you. Please suggest me more books/resources that have helped you to learn DL/NN
– GKK
Feb 22 at 13:27
• it helped me mostly by solving problems myself, especially when coding, because programming is debugging your thinking:) you learn best by doing. therefore it would be beneficial to know, what are you mostly interested in? ML/NN is a broad field, from CNN to GNN to NLP ... Feb 22 at 23:18
• "ML/NN is a broad field, from CNN to GNN to NLP" That's what I am trying to tell you. I am new and seeking advice from this forum. MY course started right of with topics like "feature descriptor", signature, fourier descriptor, intensity histogram, co occurrence matrix which I am not able to find in the recommended books. I don't know where to start. if you don't mind kindly tell me how to contact you so that I can explain my doubts.
– GKK
Feb 23 at 5:45