The principle of matrix factorization algorithm-FM model introduction

Matrix decomposition model to find two matrices, vector table hidden items and user interests
of, the scoring matrix decomposition in the form of Q and P is the product of two matrices, based on these two matrices to predict a user score of an item, Recommendation based on this score is the principle of the matrix factorization algorithm.




The introduction of FM model
Logistic regression model LR

needs to be improved-second-order cross-terms

Introduced deep learning ideas