Looks like I’ve spent my first 24h studying ML, nice.
I did a lot of math assignments regarding calculus and linear algebra. I didn’t go over too many new things but instead worked on the things I had already learned.
There’s a small topic I forgot to talk about in the post two days ago: elementary matrices (also known as elimination matrices).
Elementary matrices effectively illustrate the principle of making row operations on matrices. Take this matrix A as an example:
A=246254473
What if we wanted to eliminate a21 as the first elimination step to get the upper triangular matrix U?
Well, you can multiply A with an elementary matrix E21. Elementary matrices are derived from the identity matrix I, and their purpose is to do some slight adjustment to the matrix with row operations. Let’s see how this would go for our example.
So, we need to subtract 2⋅R1 from R2 to eliminate the element at a21. Let’s first define the elementary matrix E21:
E21=1−20010001
Notice that it indeed looks very similar to the identity matrix. The only purpose of this matrix is to just do the row operation. Let’s see how this is done with matrix multiplication:
2462544731−20010001=2062144−13
To get the upper triangular matrix, we also need to eliminate the elements a31 and a32. Let’s first define the elementary matrices for those operations: