- #1
nobben
- 2
- 0
Hi,
I'm trying to derive a result from this paper:
http://www.research.yahoo.net/files/HuKorenVolinsky-ICDM08.pdf
given cost function
[tex]
\min_{x_* , y_* } \sum_{u, i} c_{ui} (p_{ui} - x^T_u y_i)^2 + \lambda \left( \sum_u ||x_u||^2 + \sum_i ||y_||^2 \right)
[/tex]
Where both xu and yi are vectors in ℝk.
I want to find the minimum by using alternating least squares. Therefore I fix y and find the derivative with respect to xu.
cui, pui and λ are constants.
They derive the following:
[tex]x_u = \left( Y^TC^uY + \lambda I \right) ^{-1} Y^T C^u p \left( u \right)[/tex]
I'm unsure how reproduce this result...
I don't know if I should try to derive with respect to the whole vector xu or against one entry k (xuk) in the vector and then try to map this to a function for the whole vector?
Any pointers are very much appreciated.
I'm trying to derive a result from this paper:
http://www.research.yahoo.net/files/HuKorenVolinsky-ICDM08.pdf
given cost function
[tex]
\min_{x_* , y_* } \sum_{u, i} c_{ui} (p_{ui} - x^T_u y_i)^2 + \lambda \left( \sum_u ||x_u||^2 + \sum_i ||y_||^2 \right)
[/tex]
Where both xu and yi are vectors in ℝk.
I want to find the minimum by using alternating least squares. Therefore I fix y and find the derivative with respect to xu.
cui, pui and λ are constants.
They derive the following:
[tex]x_u = \left( Y^TC^uY + \lambda I \right) ^{-1} Y^T C^u p \left( u \right)[/tex]
I'm unsure how reproduce this result...
I don't know if I should try to derive with respect to the whole vector xu or against one entry k (xuk) in the vector and then try to map this to a function for the whole vector?
Any pointers are very much appreciated.