GAIN: A Gated Adaptive Feature Interaction Network for Click-Through Rate Prediction

Abstract

CTR (Click-Through Rate) prediction has attracted more and more attention from academia and industry for its significant contribution to revenue. In the last decade, learning feature interactions have become a mainstream research direction, and dozens of feature interaction-based models have been proposed for the CTR prediction task. The most common approach for existing models is to enumerate all possible feature interactions or to learn higher-order feature interactions by designing complex models. However, a simple enumeration will introduce meaningless and harmful interactions, and a complex model structure will bring a higher complexity. In this work, we propose a lightweight, yet effective model called the Gated Adaptive feature Interaction Network (GAIN). We devise a novel cross module to drop meaningless feature interactions and preserve informative ones. Our cross module consists of multiple gated units, each of which can independently learn an arbitrary-order feature interaction. We combine the cross module with a deep module into GAIN and conduct comparative experiments with state-of-the-art models on two public datasets to verify its validity. Our experimental results show that GAIN can achieve a comparable or even better performance compared to its competitors. Furthermore, in order to verify the effectiveness of the feature interactions learned by GAIN, we transfer learned interactions to other models, such as Logistic Regression (LR) and Factorization Machines (FM), and find out that their performance can be significantly improved.

GMINN: Gate‐enhanced multi‐space interaction neural networks for click‐through rate prediction

“…A multi‐channel gated mechanism (MCGM) 19 designed a hierarchical gated mechanism to select feature information at different levels in CTR predction. GAIN 20 devised a cross‐module approach to eliminate irrelevant feature interactions while preserving the informative ones. Some researchers 21,22 also employ gating mechanism to distill existing models, aiming to obtain a better CTR prediction model.…”

Feature Interaction-Based Reinforcement Learning for Tabular Anomaly Detection

“…BH-FIS implements the enumeration of all feature interactions by using outer-product and masking techniques and employs spike-and-slab priors to distinguish useful feature interactions from useless ones. Liu et al [29] propose a gated adaptive feature interaction network (GAIN) that can adaptively learn high-order feature interactions. GAIN consists of a cross-module and a deep module; the former exploits multiple parallel interaction units to explicitly model feature interactions, while the latter leverages an MLP to model feature interactions in an implicit way.…”

“…The most notable difference between tabular data and other types of data is the associative relationship between columns. Practices [26][27][28][29] in Click-Through prediction (CTR) demonstrate that feature interactions, especially high-order feature interactions, are crucial to modeling tabular data. However, to the best of our knowledge, no paper has yet worked on how to apply feature interaction to TAD.…”

“…To settle the problem that deep models do not model tabular data well, we introduce a tabular data modeling approach named gated adaptive feature interaction network (GAIN) [29]. GAIN exploits multiple parallel interaction units to learn useful high-order feature interactions.…”