What kind of knowledge is required to jump into the field of AI?

What kind of knowledge is required to jump into the field of AI? What mathematics is required? How good I should be in mathematics?

Currently, I have just started programming. I would be grateful if you suggest me reading material.

To excel in in AI you need a mathematical intuition or point of view. In order to become a full stack AI engineer, it is important that you have a firm understanding of the mathematical foundations of machine learning.

My advice to anyone preparing to jump into the field is that learning mathematics is about doing. Remember the 20/80 rule. You need to study theory 20% of the time and practice what you have learnt 80% of the time in order to be proficient.

The first step before you dive into undergraduate math is for you to refresh on foundation level math. This includes revisiting and mastering high school mathematics especially Algebra. This is necessary for you to understand higher courses.

Linear Algebra

Matrix Operations, Projections, Eigenvalues & Eigenvectors, Principal Component Analysis (PCA), Singular Value Decomposition (SVD), Eigen-decomposition of a matrix, LU Decomposition, QR Decomposition/Factorization, Symmetric Matrices, Orthogonalization & Orthonormalization, Vector Spaces and Norms are needed for understanding the optimization methods used for machine learning.

A nice thing about Linear Algebra is that there are wonderful online resources such as Khan Academy's course on linear algebra https://www.khanacademy.org/math/linear-algebra.

Probability Theory and Statistics

Probability Rules & Axioms, Conditional and Joint Distributions, Standard Distributions (Bernoulli, Binomial, Multinomial, Uniform and Gaussian), Moment Generating Functions, Bayes' Theorem, Random Variables, Variance and Expectation, Maximum Likelihood Estimation (MLE), Prior and Posterior, Maximum a Posteriori Estimation (MAP) and Sampling Methods.

I refer you to this online Statistics and Probability MOOC at Khan Academy https://www.khanacademy.org/math/probability

Algorithms and Optimization

Knowledge of data structures (Binary Trees, Hashing, Heap, Stack etc), Dynamic Programming, Randomized & Sublinear Algorithm, Graphs, Gradient/Stochastic Descents and Primal-Dual methods are needed.

For in depth tuition, I recommend that you sign up for Andrew Ng's Coursera course available here https://www.coursera.org/learn/machine-learning.

Multivariate Calculus

Topics such as Differential and Integral Calculus, Directional Gradient, Hessian, Jacobian, Laplacian and Lagragian Distribution, Partial Derivatives, Vector-Values Functions. Here is a lint to Khan Academy's Calculus course https://www.khanacademy.org/math/calculus-home.

Others

This consists of math topics that are not covered in the above four major areas. They include Information Theory (Entropy, Information Gain), Function Spaces and Manifolds, Real and Complex Analysis (Sets and Sequences, Topology, Metric Spaces, Single-Valued and Continuous Functions, Limits, Cauchy Kernel, Fourier Transforms).

More resources

Below is an Excellent resource of free Machine Learning Mathematics ebooks http://blog.paralleldots.com/data-science/list-of-free-must-read-books-for-machine-learning/

Finally to keep up with recent development and the latest papers, I recommend that you follow this blog, which aggregates AI and ML papers, http://www.arxiv-sanity.com/.

The current, cutting edge AI methods all heavily rely on statistical modeling. You might want to browse the Data Science and Cross Validated stacks to see what people are doing, and the types of maths they are using. (This is not really my field, so I'll leave it to the Neural Network and Deep Learning crowd to provide more detail here.)

I'd also strongly recommend looking at Combinatorics, Combinatorial Game Theory, and Computational Complexity Theory, since one of the the core problems of optimal decision making is intractability.

Game Theory is also important, because the other core issues of optimal decisio making involve imperfect information and incomplete information. (The former is seen more in combinatorial games, but both lead to the need for probability analysis.)

• Excellent recommendations for the Data Science and Stat sites. I will be checking them out personally, I think that is a good idea. – Seth Simba Jan 8 '18 at 18:35
• I will try this for sure – the_alonecoder Jan 10 '18 at 19:59
• @the_alonecoder thanks for the vote of confidence. The type of AI I'm interested in isn't really on the cutting edge (i.e. Deep Learning is where it's at!) but I like working in areas where there isn't much focus, and I still think there may be potential for useful extensions of classical techniques. – DukeZhou Jan 10 '18 at 20:21

Jumping into neural networks isn't that complicated.

NEURAL NETWORKS DEMYSTIFIED 1: Classification Problems helped me get started.

• Can you please follow what this research site is all about.The question and answer both should be closed. – quintumnia Jan 7 '18 at 17:10
• Looks like they were both modified to be a little bit more reasonable thanks to DukeZhou :) – Nicholas Hylands Jan 8 '18 at 21:15
• Also, there's no reason to be shutting people down when they are interested in a field of study. Even if it's in the wrong place, an encounter like that could easily dismiss interest. Snuffing out sparks should not be your primary concern. – Nicholas Hylands Jan 8 '18 at 21:16
• Don't take it personal,question like this are not inline with community guidelines.According to your analysis,does the question indicate research perspective? You would have written this as a favorite comment. – quintumnia Jan 9 '18 at 5:26
• The purpose of this board is not to just refer to other materials - it's to curate answers not found on other sites. – Avik Mohan Jan 23 '18 at 6:36

The OP asked about AI, not Machine Learning. Machine Learning is a sub-discipline of AI. Most AI used in games and motion planning does not involve machine learning.

The background you will need to "jump into the field of AI":

• Math through calculus
• Algorithms
• Basic combinatorics and probability theory

If you already have these requirements, I suggest looking at different types of AI to see what sounds most interesting to you. Mostly, getting into AI just means learning how the algorithms work. The algorithms for facial recognition (a machine learning problem) for example are much different than coordinating a team of robots (a combinatorial optimization problem).

• Thanks for contributing and welcome to AI! We've definitely had questions asking about terminology, and I strongly agree ML is a subfield of the umbrella of AI. I have noted that a very large proportion of question in the past couple of years have been focused on ML sub-fields, and it seems to be the area with the most "heat" at present. In regards to your answer, would you consider pathfinding important enough to warrant a specific mention? – DukeZhou Jan 22 '18 at 22:28
• I would say combinatoric search deserves mention -- pathfinding is a type of combinatoric search and falls under that umbrella. – thayne Jan 23 '18 at 15:52

You can start with they videos on Youtube. It is simple, with a basic knowledge in Python, you can do much things with ML.

Machine Learning - Recipes #1

I found too, a new video of Cloud AI Adventures explain "The 7 Steps of Machine Learning", you can watch HERE.

About math, I asked the same question when I started learning ML / AI. I'm not good with math, just the basics (also developer). But today, with TensorFlow and other libraries, you can focus on the theory and definition. So, do not worry, over time, AI models will be construct with a few buttons and labels ... maybe it's happening.