文章作者:覃然斌、王科 58同城 内容来源:58技术 01 系统框架 本文介绍深度学习在58商业排序CTR预估建模中的一些实践经验,主要聚焦于房产场景,从样本数据构建、模型算法迭代和线上推理优化三方面进行阐述。 02 1. 特征工程① 基础特征② 高阶特征③ Bias特征2. 样本构建:原始样本一致性与实时性03 训练样本
小结:04 算法模型 1. Wide&Deep2. DeepFM3. DIN & DIEN4. Multi-Task5. 小结05 系统工程 1. 应用层① Pipline优化② 统一特征预处理③ 模型协议修改2. 系统层加速
3. 小结06 总结与展望 覃然斌:58商业产品技术部资深算法工程师,专注于广告预估迭代与优化。 王科:58商业产品技术部-策略技术团队算法开发工程师, 主要从事58同城租房点击率预估的的优化与迭代。 07 参考资料 [1]. Mihajlo Grbovic and Haibin Cheng. Real-timePersonalization using Embeddings for Search Ranking at Airbnb. KDD '18:Proceedings of the 24th ACM SIGKDD International Conference on KnowledgeDiscovery & Data Mining (2018) [2]. Paul Covington Jay Adams Emre Sargin. Deep NeuralNetworks for YouTube Recommendations. Proceedings of the 10th ACM Conference onRecommender Systems, ACM, NewYork, NY, USA (2016) (toappear) [3]. Beutel, A., Chen, J., Doshi, T., Qian, H., Wei, L.,Wu, Y., Heldt, L., Zhao, Z., Hong, L., Chi, E. H., et al. Fairness inrecommendation ranking through pairwise comparisons. arXiv preprint. arXiv:1903.00780,2019. [4]. H.-T. Cheng, L. Koc, J. Harmsen, T. Shaked, T.Chandra, H. Aradhye, G. Anderson, G. Corrado, W. Chai, M. Ispir, et al. Wide& deep learning for recommender systems. arXiv preprint arXiv:1606.07792,2016. [5]. Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang .2017. Deep & Cross Network for Ad Click Predictions. arXiv preprintarXiv:1708.05123 (2017). [6]. W. Zhang, T. Du, and J. Wang. Deep learning overmulti-eld categorical data.In ECIR, 2016. [7]. Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li,and Xiuqiang He. 2017. DeepFM: A factorization-machine based neural network forCTR prediction. In Proceedings of the IJCAI. 2782--2788. http://dl./citation.cfm?id=3172077.3172127 [8]. Xiangnan He and Chua Tat-Seng. 2017. Neuralfactorization machines for sparse predictive analytics. In Proceedings of theSIGIR. ACM, Shinjuku, Tokyo, 355--364. [9]. Xiao Ma, Liqin Zhao, Guan Huang, Zhi Wang, Zelin Hu,Xiaoqiang Zhu, and Kun Gai. 2018. Entire Space Multi-Task Model: An EffectiveApproach for Estimating Post-Click Conversion Rate. SIGIR (2018). [10]. GuoruiZhou, Chengru Song, Xiaoqiang Zhu, Xiao Ma, Yanghui Yan, Xingya Dai, Han Zhu,Junqi Jin, Han Li, and Kun Gai . 2017. Deep interest network for click-throughrate prediction. arXiv preprint arXiv:1706.06978 (2017). [11]. G.Zhou, N. Mou, Y. Fan, Q. Pi, W. Bian, C. Zhou, X. Zhu, and K. Gai, Deep interestevolution network for click-through rate prediction. ArXiv, 2018. [12]. Q.Pi, W. Bian, G. Zhou, X. Zhu, and K. Gai. Practice on long sequential userbehavior modeling for click-through rate prediction. In KDD, pages 2671–2679, 2019. [13]. JunXiao, Hao Ye, Xiangnan He, Hanwang Zhang, Fei Wu, and Tat-Seng Chua. 2017.Attentional factorization machines: Learning the weight of feature interactionsvia attention networks. arXiv preprint arXiv:1708.04617 (2017). [14].https://software.intel.com/zh-cn/articles/lower-numerical-precision-deep-learning-inference-and-training [15].B.Jacob, S. Kligys, B. Chen, M. Zhu, M. Tang, A. Howard,H. Adam, and D.Kalenichenko. Quantization and training of neural networks for efficientinteger-arithmetic-only inference. CVPR, 2018 [16]. V.Vanhoucke, A. Senior, and M. Z. Mao. Improving the speed of neuralnetworks on cpus [C]. NIPSw, 2011. [17]. Intel(R)MKL-DNN, “Intel(R) Math Kernel Library for Deep Neural Networks. https://intel./mkl-dnn/index.html. [18].GEMMLOWP,“Gemmlowp: a small self-containedlow-precision GEMM library.” https://github.com/google/gemmlowp. [19].ARM,“Arm cmsis nn software library.”http://arm-software./CMSIS 5/NN/html/index.html. [20].Nvidia,“8 bit inference with TensorRT”. http://on-demand./gtc/2017/presentation/s7310-8-bit-inferencewith-tensorrt.pdf. [21].https://software.intel.com/en-us/frameworks/tensorflow [22].https://github.com/tensorflow/tensorflow/tree/master/tensorflow/compiler/xla [23]. https://docs./2019_R2/index.html [24].T.Chen, T. Moreau, Z. Jiang, L. Zheng, E. Yan, M. Cowan, H. Shen, L. Wang, Y. Hu,L. Ceze, C. Guestrin, A. Krishnamurthy, TVM: An automated end-to-end optimizingcompiler for deep learning,2018,[online]Available: https:///abs/1802.04799. [25]. https://yq.aliyun.com/articles/569539 [26]. YueFeng, Jun Xu, Yanyan Lan, Jiafeng Guo, Wei Zeng, and Xueqi Cheng. 2018. FromGreedy Selection to Exploratory Decision-Making: Diverse Ranking withPolicy-Value Networks. In SIGIR (SIGIR'18). 125--134. [27]. WentaoOuyang, Xiuwu Zhang, Shukui Ren, Li Li, Zhaojie Liu, Yanlong Du. Click-ThroughRate Prediction with the User Memory Network. KDD 2019 |
|