TA input may be useful to confirm what I'm saying is correct: but basically in least squares you can see sometimes we don't have a unique solution since the matrix isn't invertible - no unique solution means problem isn't convex. However when adding a L2 term we see we we modify the matrix that needs to be inverted and it becomes invertible as long as lambda >0. Unique solution means the problem was made convex. However I don't really have the rigorous proof..
Mock 2018: MCQ
Hello,
For this question in the mock exam 2018, I cannot find an example where the affirmation 1 is correct... Could you enlighten me on this ?
Cheers !!
quest.JPG
TA input may be useful to confirm what I'm saying is correct: but basically in least squares you can see sometimes we don't have a unique solution since the matrix isn't invertible - no unique solution means problem isn't convex. However when adding a L2 term we see we we modify the matrix that needs to be inverted and it becomes invertible as long as lambda >0. Unique solution means the problem was made convex. However I don't really have the rigorous proof..
Add comment