Normalize_row ord=2

Any more clarity on the difference between ord=2 vs the default ‘fro’ normalization. Not much clarity found on the internet. Why is ord=2 selected?

Hey @samuel_varghese,
A lot of mathematical information regarding this can be found on the web. Personally, I never looked into this, but let me try to mention some resources:

  • What is the difference between the Frobenius norm and the 2-norm of a matrix? | Take a look at the first answer
  • Numpy Documentation | Take a look at the examples in this doc, especially LA.norm(b) and LA.norm(b, 2). You can clearly see that they are different.
  • Now, I guess by “singular value” in the definition of 2-norm it is referring to the eigenvalues, so let me try to make a tutorial for you explaining how we can find 2-norm and Frobenius norm using simple mathematical operations. Give me some time, I will update this post soon.

Cheers,
Elemento

Hey @samuel_varghese,
Please check the out 8th version of this kernel. I have tried to implement both Frobenius-norm and 2-norm using basic numpy functions, and mentioned some of the references as well. Let me know if this helps.

Cheers,
Elemento