Machine Learning Tools or Concepts? where to start? Part 1

Few months ago, an old friend reached out asking “any ideas how to start studying AI other than python course and tensor flow” ..

Hmm … I started thinking throughly about his question, and didn’t rush in answering it, as i know he has a strong background in software engineering, and he doesn’t need a quick or basic answer. I waited for a few days before I respond back saying “That is an interesting question, and the answer can be short or long and deep based on your goals of what exactly do you like to do by studying AI or machine learning - and I started asking about some specifics, in order to tell the right way to do it, knowing that python and tensorflow in his question are just tools, useful but not for every thing!”.

My friend replied back saying: “Mainly doing Regression analysis and what’s next scenarios. It is all related to sales AI”.

And since I don’t know much about the sales field’s needs, I started to think about the technical side of the question, that guides and leads him at least to the regression analysis and to build a core knowledge in machine learning. and he will focus more on how he can connect this with his own field (sales in that case). I’ve also thought that my answer could be of good guidance for others who either have similar questions or want to know machine learning in more depth, so I decided to write this in a blog or a series of blogs. So, let’s go ahead and start, but before we start and as a disclaimer, all the points and opinions listed in this series are all my personal learning experience I acquired from and by the interactions with machine learning and data scientist working in the filed, and if you have any comments or notes, please share in the comments sections.

And here is my detailed answer:

Well .. in order to start studying ML and AI .. Python and Tensorflow and all these languages and frameworks are means and tools, having them in your toolbox might accelerate the learning and getting to your targets faster, but they are not the core …

In order to make it right. you have to start as follows:

Math for ML

Intuitively understand the mathematics and statistical concepts needed in the context of machine learning .. If you master those concepts, it will be easier for you to grasp more advanced ML concepts and topics that help you throughout your projects (sales, marketing domains or any other domain). and here is the list of topics:

1- Linear Algebra

2- Multivariate Calculus

3- Probability and Statistics, and finally

4- Numerical Optimization

Try also to understand the basic mathematics for Gradient Descent (GD) as a primer topic for ML applications.

Well, the above mentioned topics, if you searched for a reference for each topic, you will take a year to study them, and won’t finish. So, I will summarize here what you need to focus on:

• In Linear Algebra, you need to know about Vectors and Matrices, and you need to understand the concept of a vector and its notation, the geometry of vectors, and finally the Vector norms (L1, L2 and L-infinity). you also need learn the Linear Algebra Operations: like Vector inner products, Matrix products , geometric interpretations of the above, the concept of Matrix inverse and singular matrix and the Linear dependence.
• In Probability, you need to know about the outcomes and events, Rules of probability, Conditional probability, Bayes’ rule (very important rule), Random variables, Probability mass versus probability density, Expectation, and the Loss functions and maximum likelihood
• In Calculus, first you need to know about the Univariate derivatives: the Derivatives of power, exponentials, and logarithms, Leibniz and Lagrange notation, Chain rule, product rule, Minima, Maxima, local and global, Second derivatives , Gradient descent and Newton’s method. and in the Multivariate derivatives: you need to know more about Derivative of RN → R functions, Vector notation for derivatives, Matrix notation for second derivative, Multivariate minima and maxima (optimization)and saddle points, and most importantly the Concept of convexity.

Summary: in this article we covered the basic and most needed math topics needed to master the machine learning and get the right intuition out of your learning experience. In the next article, we will cover some data science related topics, that put you on the road to start understanding and building ML models that are based on a solid understating of math topics we listed here. Keep tuned for the next article.