#eth/cil/theory

Preprocessing / Normalization

make rows/columns more comparable and subtract out rating bias

Mean rating per user/item
4-archive/cil/theory/old/assets/02-mean-row-col.png
This lets us center the data:

$μ = E [X]$
$σ^{2} = Var [X]$
Then the noramlized scores or z-scores are:

Z = \frac{X - μ}{σ}