DeepFM: shared embeddings for low- and high-order feature learning

What changed

Jointly trained a factorization-machine branch and deep neural branch sharing raw-feature embeddings, combining low-order interactions and learned nonlinear features.

Scientific scope

Jointly trained a factorization-machine branch and deep neural branch sharing raw-feature embeddings, combining low-order interactions and learned nonlinear features. Inherits FM, MLP and Wide&Deep principles; not invention of them. Reduced manual cross-feature design is not elimination of all preprocessing or a guarantee that arbitrary high-order interactions are exactly learned. CTR benchmark gains do not establish causal effects or long-term user welfare.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits FM, MLP and Wide&Deep principles; not invention of them. Reduced manual cross-feature design is not elimination of all preprocessing or a guarantee that arbitrary high-order interactions are exactly learned. CTR benchmark gains do not establish causal effects or long-term user welfare.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• DeepFM: A Factorization-Machine based Neural Network for CTR Prediction

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction