RA-Internal and Contextual Attention Network for Cold-start Multi-channel Matching in Recommendation

1.运用场景

    CTR模型/item冷启动/多队列召回。

2.创新点

    highlight the cold-start channel issue in multi-channel matching,which is essential in real-world integrated personalized recommendation systems.
    first to combine the field-level internal attention with contextual attention in the matching module.

3.算法原理

3.1 网络框架

3.2 Internal and Contextual Attention Network for Cold-start Multi-channel Matching in Recommendation

    原论文

4.算法理解

    多item多通道召回，存在item有冷启动问题，将多通道信息融合做推荐。