SIGIR'22 "Microsoft" CTR estimation: using context information to promote feature representation learning

Paper title ：

Enhancing CTR Prediction with Context-Aware Feature Representation Learning

Thesis link ：

https://arxiv.org/pdf/2204.08758.pdf

code:

https://github.com/frnetnetwork/frnet

In this paper, we consider , Feature representation and context （context） The relationship between , Feature refinement network is proposed FRNet, The module learns bit levels for each feature in different contexts （bit-level） Context aware feature representation .FRNet It consists of two key components ：

• 1） Information extraction unit （IEU）, It captures contextual information and cross feature relationships , To guide the feature refinement of context awareness ;
• 2） Complementary selection gate （CSGate）, It adaptively will be in IEU The original and complementary feature representation of learning is combined with bit level weight .

FRNet It's a module , Can be compared with other ctr Model combination to improve performance . about CTR The basic process of the base model will not be repeated here , If you want to know more, you can go to the third chapter of the paper to read .