Struggling to learn Machine Learning on my own

I'm studying Machine Learning on my own, but with some difficulties. I tried many books on Machine Learning but it wasn't easy to find the right one.

Bishop's "Pattern Recognition and Machine Learning" is a very hard read. In my opinion, the problem is not the material but the exposition. Many derivations are left to the reader and there are too many "it's trivial to see", "it can be readily seen" and "after some straightforward algebra". In the end, I gave up.

"The Elements of Statistical Learning" (Hastie et al.) suffers from the same problems. Moreover, some explanations I couldn't follow because they referred to concepts I wasn't familiar with. All these books claim that you just need to know some calculus, probability and linear algebra, but that's a lie. Bishop even tries to teach you basic probability, suggesting that his book can be read by one who doesn't know probability, which is absurd.

The lessons by Andrew Ng are easy to follow but they're not very deep. They show you a collection of techniques but they don't provide the theory that should guide you in using these techniques. Also, many techniques are presented in their simplest form. The lectures by Tom Mitchell are also very easy to follow but I think he oversimplifies things. By the way, he says that after taking his course one can do research and read papers without problems. I wish it was that easy.

The same material can be presented at very different levels and I found out that there is a noticeable gap between elementary texts and lessons, and advanced ones. I was looking for something advanced but at the same time accessible. In my opinion, a text can be advanced and at the same time introductory. Introductory should mean that no prior knowledge of the topic is assumed. Many books claim to be introductory but they're not. Some explanations are so cryptic that only one with a prior exposure to the material would benefit from them.

Finally, I found the right book for me: Pattern Classification (Duda, Hart, Stork, 2ed.). The book doesn't shy away from advanced material and the explanations are great. Finally, a book that I can study on my own without having to rely on somebody else for additional explanations!

What's your experience with Machine Learning books, textbooks and lectures?