Overview of matrix factorization

Créer des moteurs de recommandation avec PySpark

Jamen Long

Data Scientist at Nike

Matrix Factorization

one matrix being split into two factor matrices

factor matrices multiplied to produce approximated original matrix

square matrix with positive values

matrix with two factor matrices where some of the values in factor matrices are negative

matrix and two factor matrices where all values are positive

factor matrices with highlights on latent dimensions of 3 rows or 3 columns

matrix with most cells blank

matrix with most cells blank and two factor matrices will all cells full

partially blank matrix with circle around one cell and respective factor matrix row and column highlighted

partially blank matrix with circle around one cell and respective factor matrix row and column highlighted

partially blank matrix with circle around one cell and respective factor matrix row and column highlighted

partially blank matrix with circle around one cell and respective factor matrix row and column highlighted

partially blank matrix with factor matrices and product matrix with all cells filled in

Créer des moteurs de recommandation avec PySpark