#Paper Reading# Deep Learning Recommendation Model for Personalization and Recommendation Systems

论文题目: Deep Learning Recommendation Model for Personalization and Recommendation Systems
论文地址: https://arxiv.org/abs/1906.00091
论文发表于: arxiv 2019

论文大体内容：
本文主要提出了deep learning recommendation model（DLRM）的模型，来使用pytorch进行分布式训练，效果也达到state-of-art；

Motivation：
该模型主要是用于推广pytorch框架，并开源给大家解决推荐引擎的挑战，工业届可方便使用。

Contribution：
本文作者提出的DLRM模型，能够达到轻量级和速度快，便于工业届应用。

1. DLRM模型结构如下图，主要是将稀疏特征用embedding，如id类；dense特征用MLP，如统计特征；
#Paper Reading# Deep Learning Recommendation Model for Personalization and Recommendation Systems

2. 相比DeepFM、xDeepFM等模型，DLRM模型能够减少模型的维数；
DLRM specifically interacts embeddings in a structured way that mimics factorization machines to significantly reduce the dimensionality of the model by only considering cross-terms produced by the dot-product between pairs of embeddings in the final MLP.

3. 本文还设计了一种并行化MLP的方式使得模型更快；
#Paper Reading# Deep Learning Recommendation Model for Personalization and Recommendation Systems

实验
3. Dataset
Criteo Ad Kaggle数据集[1]，是Kaggle比赛的数据集；

4. 实验结果
#Paper Reading# Deep Learning Recommendation Model for Personalization and Recommendation Systems

参考资料：
[1] https://www.kaggle.com/c/criteo-display-ad-challenge

以上均为个人见解，因本人水平有限，如发现有所错漏，敬请指出，谢谢！

#Paper Reading# Deep Learning Recommendation Model for Personalization and Recommendation Systems

相关推荐