1 views collected around this technical thread.
This article reviews the latest academic advances in multimodal recommendation systems, covering background, system workflow, modal encoders, feature interaction (connection, fusion, filtering), feature enhancement, model optimization, and future research challenges.
This article presents a memory‑driven architecture (HCNet and MemoNet) that equips recommendation models with scaling‑law characteristics by storing and retrieving arbitrary feature‑combination embeddings, evaluates multi‑hash codebooks, memory‑restoring strategies, key‑feature selection, and demonstrates significant offline and online performance gains.
This article reviews the development of CTR prediction models from early collaborative‑filtering methods to modern deep‑learning approaches, discusses core challenges such as feature interaction and continuous‑feature embedding, introduces recent Huawei solutions like AutoDis and ScaleFreeCTR for efficient large‑embedding training, and outlines future research directions.
This article reviews the evolution of recommendation models from early collaborative filtering to modern deep learning approaches, discusses core challenges such as CTR prediction, outlines user‑behavior and combination‑feature modeling techniques, introduces large‑embedding training and continuous‑feature embedding methods like AutoDis, and presents distributed training frameworks such as ScaleFreeCTR, concluding with future research directions.
The Co‑Action Network replaces costly Cartesian‑product feature crossing with lightweight micro‑net‑based interaction units that share parameters across feature pairs, delivering comparable CTR prediction accuracy while cutting parameters to one‑tenth and boosting online latency, as proven in large‑scale advertising deployments.
This article systematically reviews the evolution of click‑through‑rate (CTR) prediction models—from early distributed linear models like logistic regression, through automated feature engineering with GBDT+LR, various factorization‑machine variants, embedding‑MLP shallow modifications, dual‑tower combinations, and advanced explicit feature‑cross networks—highlighting each model’s structure, advantages, limitations, and comparative insights.
This article details the development and deployment of CAN (Co‑Action Net), a novel click‑through‑rate prediction model that captures item‑item co‑action via attention‑based slot embeddings, offering superior performance to Cartesian‑product methods while reducing parameter and serving costs.
This article explains how feature engineering transforms raw data into numerical representations for machine‑learning models, covering mapping of numeric and categorical values, one‑hot and multi‑hot encoding, sparse representations, scaling, handling outliers, binning, data quality checks, and feature interactions to capture non‑linear relationships.
This article introduces the Deep Sparse Network (NON), a new deep neural architecture for recommendation systems that combines field‑wise networks, across‑field interaction networks, and an operation‑fusion network, and demonstrates its superior performance through extensive experiments and ablation studies.
This article describes iQIYI’s dual‑DNN ranking architecture that combines a high‑capacity teacher network with a lightweight student network via online knowledge distillation, addressing the trade‑off between model effectiveness and inference efficiency in large‑scale recommendation systems.
iQIYI’s dual‑DNN ranking model uses an online teacher‑student knowledge‑distillation framework where a complex teacher DNN shares representations with a lightweight student DNN, enabling end‑to‑end training, large‑scale feature crossing, and substantially higher recommendation accuracy while cutting inference latency and model size.