Contents
What is factorization machine?
Factorization Machines (FM) are generic supervised learning models that map arbitrary real-valued features into a low-dimensional latent factor space and can be applied naturally to a wide variety of prediction tasks including regression, classification, and ranking.
What is matrix factorization model?
Matrix factorization is a class of collaborative filtering algorithms used in recommender systems. Matrix factorization algorithms work by decomposing the user-item interaction matrix into the product of two lower dimensionality rectangular matrices.
Which is field aware factorization machine for CTR prediction?
Recently, a variant of FMs, field-aware factorization machines (FFMs), outperforms existing models in some world-wide CTR-prediction competitions. Based on our experiences in winning two of them, in this paper we establish FFMs as an effective method for classifying large sparse data including those from CTR prediction.
What are the parameters of Feild factorization machine?
The FM factorization of Q has K × l parameters. The “feild-aware” FM has K × l × | Z | parameters. A model with all interactions has K × ( K + 1) / 2 parameters. Standard factorization machines have fields too. The “novelty” here seems to be the use of GBDT features and the application of the hashing tricks.
Is the factorization process an improved version of MF?
The factorization process has to learn all these from existing interactions. Hence, factorization machines are introduced as an improved version of MF. ( Since this article is focused on FFM, I will not delve into greater details of MF. To find out more, I highly recommend Google’s introductory course on recommender systems.)
How are factorization machines used in sparse settings?
Broadly speaking, factorization machines are able to estimate interactions in sparse settings because they break the independence of the interaction parameters by factorizing them (using latent vectors as expressed in ).