# Why learn Math for AI Many students learning AI on their own will rely on the short review of Math and Probability and Statistics often given in the first few chapters or appendix of an ML textook or worse DL textbook. Here, I share some insight into why this is a mistake. ## Prerequisites Here are some undergraduate topics/courses that you should know before learning AI/ML: - Calculus - Discrete Mathematics - Advanced Linear Algebra - Probability and Statistics - Statistical Programming The following courses are usually required for a computer science degree: - Algorithms and Data Structures - Computer Organization - Computer Networks - Operating Systems ## Discussion As you can see by the list above, it will take a student a lot of time to just read a textbook on all those subjects. Therefore, I am confident that a vast majority of students learning independently will choose to skip a detailed study of the subjects. Since I have an M.S. in Mathematics and MSCS in Artificial Intelligence I have a unique perspective. I quickly discovered that my studies in mathematics were well spent. If we browse the table of contents of a textbook on **Discrete Mathematics**, you find that pretty much all the chapters cover material that is discussed in graduate courses on AI and ML. For **Computer Organization**, we find that a knowledge of how computers work is essential for memory optimization and tuning which becomes important as AI moves toward IoT and Robotics. In fact, I was recently amazed when a programmer tried to argue that a GPU somehow magically processes and uses memory differently from the CPU and RAM with which it interfaces which are basic computer organization concepts. **Probability and Statistics** are essential in the analysis and preprocessing of the data required for training AI models as well as many areas of AI theory and algorithms. In fact, every ML project is an statistical experiment or simulation. Most students of AI know that Linear Algebra is important but they fail to realize that neural networks are actually solving a nonlinear system of differential equations which requires knowledge of Calculus. We also find that Robotics relies heavily on methods and techniques of **Advanced Linear Algebra** that are not covered in an "Intro to Linear Algebra" course. Hopefully, all students of AI know the importance of spending time learning **Algorithms and Data Structures** which are essential to all software engineering projects including AI. Finally, this brings us to **Computer Networks** and **Operating Systems**. A network connection is essential to any modern AI system and the growth of AI in WiFi, Robotics, and IoT will make knowledge of Computer Networks and Operating System ubiquitous. Thus, why spend all this time in self-study with nothing to show for this huge accomplishment? With 85% of AI projects failing and what seems to be a lack of an ethical compass by many AI researchers, I think the pendulum is beginning to swing back toward academic credentials and certifications. Anyway, just my two cents. Enjoy!