Learning Math for Machine Learning
I have just completed the final exam of the course on Probability theory at MIT OCW. Over the span of around 2 years, this is the 4th math class that I took from MIT OCW. Earlier ones are listed below. All four courses are the fundamental mathematics courses that one needs to master before diving into machine learning. The other 2 very useful math courses relevant to Machine Learning are Statistics and Optimization.
The list of courses that I took from MIT OCW is the following:
1. Single variable calculus: see the link: bit.ly/3i8aBpT
2. Linear Algebra: see the link: bit.ly/2VtlTey
3. Multivariable Calculus: see the link: bit.ly/31qyWS4
4. Probabilistic Systems Analysis and Applied Probability: see the link: bit.ly/3eJNkbJ
Even though I took some variants of those courses during my undergraduate studies, I eventually decided to retake those courses to deepen my math for being able 1) to take graduate-level machine learning, engineering and other computer science courses and 2) to do research. Although I enjoy doing math, from now on I am excited to spend my time more on studying machine learning theory, computer science courses and working on projects.
Having said that I would like to share with you some takeaways from my math journey and self-study.
1) Having spent some time on math courses on Coursera (by completing Math for Machine Learning specialization by Imperial College and browsing the content of some other courses), I would say if you really want to get proper mathematical foundation do not waste your time doing 4–5 week courses with few exercises, take proper college-level courses with lots of exercises. For example, MIT OCW courses are exactly MIT math courses that are offered on campus, containing very high-quality lecture videos from classes, lots of challenging problem sets with solutions, recitation videos and (midterm, final) exams.
2) It is more effective and efficient to dedicate your time and once and properly learn the subject rather than doing it in a hurry, looking for shortcuts and feeling frustrated and going back to study the same subjects again.
3) Do not just watch the lectures and fall into the trap of understanding. Try to apply the learned concepts immediately. Take your time to solve problems. It is really when you solve challenging problems that require the understanding of some nuances in the subject, you truly learn (although you struggle quite a bit and it takes time).
4) From time to time (e.g. once a week) stop learning new content and go back and review the learned material and test your knowledge by taking exams/quizzes. You may feel like you’d better go forward and rather finish the course more quickly, but you will forget most of the learned material if you do not do it.
5) Basically, the more time you spend on learning the subject, the more exercises you do, the longer you will retain the gained knowledge.
6) If you decide to take courses from MIT OCW, take the SC (scholar) version of courses that are specifically designed for independent learners (it means they have video lectures, solutions to problem sets and recitation videos).
#math #calculus #linearalgebra #probability #machine #learning #data #science #online #education
Hope this helps! Feel free to share your learning experience and suggest the learning resources that you think are worth sharing.
Published By
Originally published at https://www.linkedin.com.
