Now let me say a few words about the pre-requisites to this specialization. First let me start with the programming part. I assume that you know at least some Python, and have used or at least have seen Jupyter notebooks. If you are not familiar with Jupyter notebooks, take a look at tutorial reference for you in this week's reading list. I also assume that you're familiar with other Python libraries such as numpy and Pandas. And prior knowledge of TensorFlow is not assumed, as TensorFlow will be gradually introduced within the course. Now let's talk about pre-requisites for the math part. Here is what I expect you to know. First, machine learning uses lots of linear algebra. So I expect you to be familiar with linear matrix equations, eigenvalue decomposition, inverse matrices, and other related concepts. I also assume that you know basic probability theory. For example, you're familiar with the Gaussian, exponential, or binomial distributions. Basic probability rules such as the base lemma, and you also know some basic statistics. Finally on the math side, I assume that you know basic calculus, including in particular rules of differentiation of composite functions. So that formulas like this, or like this, or even like that would not perplex you. If they do, please refresh your knowledge of calculus. As one of my heroes in science used to say, the math should not stand between you and the problem you want to solve. Just the opposite, it should be able to help you once you know how to use it. By the way, the question of how much of math that you need to know to do machine learning is a popular topic on various discussion boards. Recently I came across a very lovely post written by an ex-physicist currently working in the machine learning space. I strongly recommend you read this post, and here's the link for your convenience. And on my side, I can confirm that everything this guy says is exactly how physicists approach problems on the mathematical side. If I had to condense this just to in a few sentences, I would put it as follows. When you come across a new machine learning model, be it in our lectures, books, or original papers, start with an abstract or whatever replaces the abstract. If the statement about what the model does attracts your interest so that you want to learn more, skim through the many equations of the paper. And make sure you understand what they mean, not yet how they obtained. Do some sort of a meditation on main formulas. Observe what quantity stands in the left-hand side of the equation, and what terms appear on the right-hand side. How do they enter, for example? Exponentially or logarithmically? Then assume that all equations are right, and that an implementation available for you is right as well. And proceed directly to play hands-on with the model. You may want to first feed your model with the data you understand. For example, with purely random data, or even constant data, to see if it passes some sort of sanity checks. If it does, then feed the model your actual data you want to explore. And when you get the results, chances are that you will either like them or dislike them. But most likely, you will notice some particular behavior, and you will have questions about that. And it's only then that you can return to the main section and read the math. Now long story short, if you start with the math, just move on and come back to it later if needed. That would be my practical advise for you both for this course and beyond, unless of course you have an unlimited time budget. But if you don't, which is the most often the case in real life, the approach that I just described can save you lots of time. Finally, let’s talk about pre-requisites on finance. This will be really short, because in fact I don't assume that you have any specific knowledge in finance. And all financial concepts or problems discussed in this specialization will be properly explained for non-specialists. Okay, so I covered most of what I wanted to say here as a way of a general introduction to this specialization. And now let me conclude with the list of recommended literature. There is a number of excellent textbooks on machine learning, but there are no textbooks specifically on machine learning in finance. So what I did for this course is combining multiple sources, including in particular parts of books by Bishop, Murphy, Goodfellow, and also a very recent book by Geron. A few other books that I like a lot, the books by Marsland and another book by Gershenfeld. And in addition to textbooks, I have used original publications, my own research, industry papers, blogs, Wikipedia, posts on discussion forums, and so on. In short, any sort of digital information that I found useful for the purpose of creation of this course. As a rule, I always refer to original source whenever I based any substantial part of a lecture on any single source so that you can always look it up for more details. Also in such cases I usually keep the notation of the original publication, or adjust it only a little bit to leverage more common conventions. So if you need, you can always come back and pick it up from where I left it in the lecture. Okay, so I think this is finally all I have to say in this introductory part to this specialization. And in the next video, we will start our first course. I hope you will find it helpful and interesting, but if you feel that I move too slow, or too fast, too deep or too shallow, do not cover some topics. Or the opposite, stay too long with some other non-interesting topics, and so on, please share your thoughts on the course forum. Good luck with the course, and remember, machine learning starts here. See you soon.
