Difficulties with solving Absolute Loss LP

[jeziddah]

Hello,
I am following a course on Machine Learning: Linear Regression taught some years ago and I came across the attached problem. I have tried to implement the solution in MATLAB using both the attached MATLAB and YALMIP codes keep getting error messages regarding the matrix dimension. Is anyone here familiar with this problem and would be kind to offer some guidance?

[johanlofberg]

Huh? There are no errors showing in your pictures.

solvesdp is obsolete. The command is called optimize

[johanlofberg]

and don't use sum(abs(x)), much faster in YALMIP to use norm(x,1)

[jeziddah]

You've concluded that using n=1 is wrong, and the correct value should be 2, but you still say you don't know what to do?

I tried making theta as the same length as phi but i still get the same error

[johanlofberg]

No you don't, you are doing something else. If theta has length 2, the multiplication Phi * theta cannot fail, and neither can Phi * theta-y

[jeziddah]

I am following the solution provided in the lecture as attached to this discussion. however, I noticed that by implementing the solution to the problem, the matrix and the vector dimensions are not related as I attached earlier. So I can understand the error message. Since its a new area I am venturing into, I am struggling to figure it out. the more I search about it, the more I come across materials with the same solution that returns the attached error message.

[johanlofberg]

But as you've already concluded, the computation of n in the code is incorrect as it is based on the wrong matrix.