Hi, I am currently looking to pursue a career in Artificial Intelligence (subfields such as natural language processing, computer vision, and machine learning) and I was recently doing some research on the type of math/statistics needed for AI. I mostly came across answers saying that Calculus, Linear Algebra, Discrete Math, and Probability/Statistics (the answers did not mention specific stats concepts that were relevant) were the most useful types of math for the field. I have taken Calculus I and II, am currently taking Calculus III, and will be taking Linear Algebra and Discrete Math during the spring semester. I haven’t taken Probability/Statistics yet. What are the most important specific Probability/Statistics concepts/classes I should be learning to prepare for a career in AI? Should I start off with basic introductory Statistics or do you think it would be fine to jump straight into Calculus-based Statistics? I want to learn Probability/Statistics with computer software packages/programming languages like R, Scala, and Python. Also, what kind of math is useful/important Math besides Calculus, Linear Algebra, Discrete Math, and Probability/Statistics for AI?
It would be also be very helpful if you could give reasons for why specific types of math/statistics are useful and examples of use cases.
Probability and Statistics are very important for AI and Machine Learning. In fact a lot of AI and Machine Learning is, arguably, just applied Probability/Statistics. You should take more theory-focused courses if you think you can handle that, and you should take a Probability course and a Statistics course separately (since they are separate, but related, topics). You should start with Probability and then take Statistics.
So do you think that one semester of probability and one semester of statistics would cover that base?
Agree with Adamah. I would make sure your probability and statistics courses both require multi variable calculus and linear algebra. If they don’t, you want to find more advanced classes that are more rigorous. Many statistics classes are much too elementary to do you much good.
Some aspects of graph theory would likely be useful. Also, many schools have a numerical matrix algorithms class well beyond linear algebra that would be very useful. Algorithms is useful. Also convex optimization.
I agree with all the above writers, especially @Adamah and @ClassicRockerDad. I think it’s also important to take as many continuous math and physics courses you can. In fact, it’s fine to be a Physics or Math major and just take the machine learning electives offered at your school.
From the CS side, it’s useful to be a very good programmer and to know something about the theory of algorithms.
From the math side, calculus, linear algebra, real analysis, linear programming, convex optimization, and perhaps functional analysis (if you get so far, I didn’t).
From the physics side, quantum mechanics and statistical mechanics are very useful. I haven’t taken more advanced physics courses than that, so I’m not sure to what extent they are useful (though I think they can’t hurt). Quantum mechanics forces you to think probabilitistically. And many of the computational techniques in machine learning as well as concepts such as entropy come from statistical physics (in their CS form, it’s called Information Theory - it’s a beautiful course).
I’d suggest getting some practical experience with data mining, through competitions or for fun. This will make you more job-ready.