How ALS alternates to generate predictions

Créer des moteurs de recommandation avec PySpark

Jamen Long

Data Scientist at Nike

partially blank matrix

partially blank matrix with two factor matrices completely full

partially blank matrix with two factor matrices completely full on factor matrix has "adjusted" written over it, other matrices have "constant" written over them

same image as above with graph containing iterations on x axis and RMSE on y axis and a dot high on the left

same image as above with graph containing iterations on x axis and RMSE on y axis and a new dot lower on the y axis and further right on x axis, a line connecting the dots

same image as above with graph containing iterations on x axis and RMSE on y axis and another dot lower on the y axis and further right on x axis, a line connecting the dots

same image as above with more dots indicating the RMSE decreasing as the model continues to iterate

partially blank matrix with arrow leading to completely full approximation of other matrix

partially blank matrix

partially blank matrix with two completely full factor matrices

partially blank matrix with one value highlightd and two factor matrices with respective row and column highlighted

partially blank matrix with another value highlightd and two factor matrices with respective row and column highlighted

partially blank matrix with horizontal arrows pointing to the first value in each row and two factor matrices

partially blank matrix with vertical, downward arrows pointing to the first value in each column and two factor matrices

original matrix with two factor matrices with all values in factor matrices highlighted

original matrix with two factor matrices with all values in factor matrices highlighted with blank spaces in original matrix highlighted

fourth matrix added showing whre ALS filled in the blank cells

Let's practice!

Créer des moteurs de recommandation avec PySpark