首页 > 资料专栏 > 营销 > 营销策划 > 广告策划 > 基于向量的召回算法及其在个性化广告新闻中的应用实践-刘政PDF

基于向量的召回算法及其在个性化广告新闻中的应用实践-刘政PDF

wangsha***
V 实名认证
内容提供者
热门搜索
召回 算法
资料大小:2936KB(压缩后)
文档格式:PDF
资料语言:中文版/英文版/日文版
解压密码:m448
更新时间:2023/6/14(发布于辽宁)

类型:积分资料
积分:15分 (VIP无积分限制)
推荐:升级会员

   点此下载 ==>> 点击下载文档


“基于向量的召回算法及其在个性化广告新闻中的应用实践-刘政PDF”第1页图片 “基于向量的召回算法及其在个性化广告新闻中的应用实践-刘政PDF”第2页图片 图片预览结束,如需查阅完整内容,请下载文档!
文本描述
Embedding Based Recall: Practice,
Progress and Perspectives
Zheng Liu, Jianxun Lian, Xing Xie
Social Computing Group, MSRA
Aug 15th, 2021
Reinforced Anchor Knowledge Graph Generation for News Recommendation Reasoning, Liu et. al.
KDD 2021Outline
o Overview
o Multi-Stage Pipeline
o EBR: Pros and Cons
o Embedding learning algorithms
o Negative Augmentation
o Hard Negative Sampling
o Diversified representation
o Training as knowledge distillation
o Things beyond learning algorithms
o Efficiency issues
o Combo of sparse and denseOverview: Multi-Stage Pipeline
L1 Stage
(Recall)L2 Stage
(Rank)L3 Stage
(Re-rank)
0~1
× 106~9× 101~3 × 10
o Rank: high-precision, KPI-
oriented
o Recall: fast, accurate, comprehensive Overview: Multi-Stage Pipeline
Fast, Accurate
Ads
Query Xbox 360 4GB Slim
Console
Microsoft Xbox 360
Microsoft Xbox 360 E
250GB
Xbox 360 Game System
HDMI