.png)
Mathematics for Machine Learning
Mathematics is too vast a subject to be considered for this question. The breadth and depth of mathematical awareness you require for machine learning totally depends on what you are learning in the subject. Keeping this in mind, let's deal with what you need to know in "mathematics" for machine learning.
1. Probability and mathematical statistics This is a fundamental requirement for machine learning and so you need to know well. When I say probability it's more than what you studied in High school and almost everything you probably not paid attention to during your undergrad. You need to know about Random variables, their distributions, probabilistic convergence, and estimation theory. That covers a major part of what you need to know here.
Two of my favorite resources are:-
1. Joseph Blitzstein - Harvard Stat 110 lectures
2. Larry Wasserman's book - All of statistics
2. Linear algebra
Linear algebra will pop up every now and then in ML. PCA, SVD, LU decomposition, QR decomposition, symmetric matrices, orthogonalization, projections, matrix operations are needed many times. The good thing is that there are countless resources available on linear algebra.
My all-time favorite is Gilbert Strang's MIT lectures on linear algebra.
3. Optimisation
Though only a few things from optimization are needed most of the time, a strong foundational knowledge will help a long way. You need to know Lagrange multipliers, gradient descent, and primal-dual formulation. The best resource on this is Boyd and Vandenberghe's course on Convex optimization from Stanford.
4. Calculus
I wanted to put this on the top, but I'm putting it in the last just to emphasize on the fact that only a fundamental knowledge is needed in terms of calculus. Know about 3-D geometry, integration, and differentiation and you'll survive. It's the easiest to start with amongst the topic I've mentioned here. MIT has good lectures on calculus.
I think with these 4 tools you'll most likely find ML easy to understand. Other than these you may find real analysis and functional analysis relevant too, but they are just formal generalizations of the topics mentioned before.
Note:- This article is taken from Quora.
1. Probability and mathematical statistics This is a fundamental requirement for machine learning and so you need to know well. When I say probability it's more than what you studied in High school and almost everything you probably not paid attention to during your undergrad. You need to know about Random variables, their distributions, probabilistic convergence, and estimation theory. That covers a major part of what you need to know here.
Two of my favorite resources are:-
1. Joseph Blitzstein - Harvard Stat 110 lectures
2. Larry Wasserman's book - All of statistics
2. Linear algebra
Linear algebra will pop up every now and then in ML. PCA, SVD, LU decomposition, QR decomposition, symmetric matrices, orthogonalization, projections, matrix operations are needed many times. The good thing is that there are countless resources available on linear algebra.
My all-time favorite is Gilbert Strang's MIT lectures on linear algebra.
3. Optimisation
Though only a few things from optimization are needed most of the time, a strong foundational knowledge will help a long way. You need to know Lagrange multipliers, gradient descent, and primal-dual formulation. The best resource on this is Boyd and Vandenberghe's course on Convex optimization from Stanford.
4. Calculus
I wanted to put this on the top, but I'm putting it in the last just to emphasize on the fact that only a fundamental knowledge is needed in terms of calculus. Know about 3-D geometry, integration, and differentiation and you'll survive. It's the easiest to start with amongst the topic I've mentioned here. MIT has good lectures on calculus.
I think with these 4 tools you'll most likely find ML easy to understand. Other than these you may find real analysis and functional analysis relevant too, but they are just formal generalizations of the topics mentioned before.
Note:- This article is taken from Quora.
