It seems a proof requires very strong probability background. But now I know at least that

to be "fixed" before starting the proof. I will defer the proof until I will reach the required math level, and will refresh my probability knowledge reserving an eye on the guidelines you send. Before I write this reply, I've reviewed some probability books that I've possess. I Could find the Markov's inequality, and Chernoff bound in some of them. I could not find Hoeffding's bound in any of them. Thus, I especially interested if you have a probability book (tailored for ML

) you can recommend. The main difficulty in ML is the level of required math and I think it is never enough.