当前位置：网站首页>SIGIR'22「微软」CTR估计：利用上下文信息促进特征表征学习
SIGIR'22「微软」CTR估计：利用上下文信息促进特征表征学习
2022-04-23 20:04:00 【智源社区】
论文标题：
Enhancing CTR Prediction with Context-Aware Feature Representation Learning
论文链接：
https://arxiv.org/pdf/2204.08758.pdf
code:
https://github.com/frnetnetwork/frnet
本文考虑点击率预估中，特征表征与上下文（context）之间的关系，提出特征细化网路FRNet，该模块在不同上下文中为每个特征学习位级别（bit-level）的上下文感知特征表征。FRNet 由两个关键组件组成：
-
1）信息提取单元（IEU），它捕获上下文信息和交叉特征关系，以指导上下文感知的特征细化； -
2）互补选择门（CSGate），它自适应地将在 IEU 中学习的原始和互补特征表征与位级权重相结合。
FRNet是一个模块，可以与其他ctr模型结合来提升性能。对于CTR的基模型的基础流程这里不再赘述，想了解的小伙伴可以前往论文的第三章阅读。
版权声明
本文为[智源社区]所创，转载请带上原文链接，感谢
https://hub.baai.ac.cn/views/16624
边栏推荐
- OpenHarmony开源开发者成长计划，寻找改变世界的开源新生力！
- MySQL advanced lock - overview of MySQL locks and classification of MySQL locks: global lock (data backup), table level lock (table shared read lock, table exclusive write lock, metadata lock and inte
- MFC obtains local IP (used more in network communication)
- 山东大学软件学院项目实训-创新实训-网络安全靶场实验平台（六）
- How to select the third-party package of golang
- MySQL syntax collation (5) -- functions, stored procedures and triggers
- 程序设计语言基础(2)
- Openharmony open source developer growth plan, looking for new open source forces that change the world!
- Lpc1768 optimization comparison of delay time and different levels
- Is meituan, a profit-making company with zero foundation, hungry? Coupon CPS applet (with source code)
猜你喜欢
The most detailed network counting experiment in history (2) -- rip experiment of layer 3 switch
MySQL syntax collation (3)
JVM的类加载过程
Esp8266 - beginner level Chapter 1
C6748 软件仿真和硬件测试 ---附详细FFT硬件测量时间
@MapperScan与@Mapper
基于pytorch搭建GoogleNet神经网络用于花类识别
Zero base to build profit taking away CPS platform official account
MySQL syntax collation (4)
Grafana shares links with variable parameters
随机推荐
filebeat、logstash配置安装
Hot reload debugging
Lottery applet, mother no longer have to worry about who does the dishes (assign tasks), so easy
渤海期货这家公司怎么样。期货开户办理安全？
Steps to build a deep learning environment GPU
C语言的十六进制printf为何输出有时候输出带0xFF有时没有
Audio editing generation software
Speex维纳滤波与超几何分布的改写
Project training of Software College of Shandong University - Innovation Training - network security shooting range experimental platform (6)
Electron入门教程4 —— 切换应用的主题
Database query - course selection system
Prefer composition to inheritance
How to select the third-party package of golang
山东大学软件学院项目实训-创新实训-网络安全靶场实验平台（五）
Efficient serial port cyclic buffer receiving processing idea and code 2
Go three ways to copy files
Easy mock local deployment (you need to experience three times in a crowded time. Li Zao will do the same as me. Love is like a festival mock)
Mfcc: Mel frequency cepstrum coefficient calculation of perceived frequency and actual frequency conversion
Go modules daily use
Build intelligent garbage classification applet based on Zero