Ruiming Tang2, Jiajin Li1, Jinkai Yu2, Huifeng Guo3, Xiuqiang He4, Shengyu Zhang1,5 1The Chinese University of Hong Kong, 2Noah’s Ark Lab, Huawei, 3Shenzhen Graduate School, Harbin Institute of Technology, 4Data service center, MIG, Tencent, 5Tencent Quantum Lab, Tencent Presented by Shunsuke KITADA Gunosy DM in Gunosy Inc. Dec 6, 2018
Yu, Jinkai and Guo, Huifeng and He, Xiuqiang and Zhang, Shengyu "Field-aware Probabilistic Embedding Neural Network for CTR Prediction" Proceedings of the 12th ACM Conference on Recommender Systems (RecSys) 2018 https://dl.acm.org/citation.cfm?id=3240396 3
youtube recommendations." Proceedings of the 10th ACM Conference on Recommender Systems. ACM, 2016. [Lie+ 2015] Liu, Qiang et al. "A convolutional click prediction model." Proceedings of the 24th ACM International on Conference on Information and Knowledge Management. ACM, 2015. [Qu+ 2016] Qu, Yanru et al.. “Product-based neural networks for user response prediction." Data Mining (ICDM), 2016 IEEE 16th International Conference on. IEEE, 2016. [Shan+ 2016] Shan, Ying, et al. "Deep crossing: Web-scale modeling without manually crafted combinatorial features." Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2016. [Xiao+ 2017] Xiao, Jun, et al. "Attentional factorization machines: Learning the weight of feature interactions via attention networks." arXiv preprint arXiv:1708.04617 (2017). [Zhang+ 2016] Zhang, Weinan et al. "Deep learning over multi-field categorical data." European conference on information retrieval. Springer, Cham, 2016. [Zhou+ 2018] Zhou, Guorui, et al. "Deep interest network for click-through rate prediction." Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM, 2018. [Cheng+ 2016] Cheng, Heng-Tze, et al. "Wide & deep learning for recommender systems." Proceedings of the 1st Workshop on Deep Learning for Recommender Systems. ACM, 2016. [Guo+ 2017] Guo, Huifeng, et al. "DeepFM: a factorization-machine based neural network for CTR prediction." Proceedings of the 26th International Joint Conference on Artificial Intelligence. AAAI Press, 2017. Introduction • 近年提案されているCTR予測モデル ◦ 深層学習ベースの手法 [Convington+ 2016; Liu+ 2015; Shan+ 2016; Xiao+ 2017; Zhang+ 2016; Zhou+ 2018] ✗ 偏った高次元特徴の相互作用を学習してしまったり シンプルな低次元特徴を捉えられなかったりする ◦ 高次元特徴・低次元特徴の相互作用を捉えるモデル ▪ Wide & Deep [Cheng+ 2016] ▪ Deep FM [Guo+ 2017] 6 Introduction > Model > Experiments > Conclusion ✗ CTR予測に有効なフィールド情報を有効活用できていない
◦ 3つのコンポーネントで 多様な特徴を学習 ▪ Linear term (LN) ▪ Quadratic term (QDR) ▪ Deep NN term (DNN) ◦ 埋め込みを確率分布とした ときの学習手法の適用 ▪ Reparameterization trick [Kingma+ 2013; Ruiz+ 2016] 9 Introduction > Model > Experiments > Conclusion [Kingma+ 2013] Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). [Ruiz+ 2016] Ruiz, Francisco R., Michalis Titsias RC AUEB, and David Blei. "The generalized reparameterization gradient." Advances in neural information processing systems. 2016.
◦ 3つのコンポーネントで 多様な特徴を学習 ▪ Linear term (LN) ▪ Quadratic term (QDR) ▪ Deep NN term (DNN) ◦ 埋め込みを確率分布とした ときの学習手法の適用 ▪ Reparameterization trick [Kingma+ 2013; Ruiz+ 2016] 10 Introduction > Model > Experiments > Conclusion [Kingma+ 2013] Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). [Ruiz+ 2016] Ruiz, Francisco R., Michalis Titsias RC AUEB, and David Blei. "The generalized reparameterization gradient." Advances in neural information processing systems. 2016.
◦ 3つのコンポーネントで 多様な特徴を学習 ▪ Linear term (LN) ▪ Quadratic term (QDR) ▪ Deep NN term (DNN) ◦ 埋め込みを確率分布とした ときの学習手法の適用 ▪ Reparameterization trick [Kingma+ 2013; Ruiz+ 2016] 14 Introduction > Model > Experiments > Conclusion [Kingma+ 2013] Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). [Ruiz+ 2016] Ruiz, Francisco R., Michalis Titsias RC AUEB, and David Blei. "The generalized reparameterization gradient." Advances in neural information processing systems. 2016.
◦ 3つのコンポーネントで 多様な特徴を学習 ▪ Linear term (LN) ▪ Quadratic term (QDR) ▪ Deep NN term (DNN) ◦ 埋め込みを確率分布とした ときの学習手法の適用 ▪ Reparameterization trick [Kingma+ 2013; Ruiz+ 2016] 15 Introduction > Model > Experiments > Conclusion [Kingma+ 2013] Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). [Ruiz+ 2016] Ruiz, Francisco R., Michalis Titsias RC AUEB, and David Blei. "The generalized reparameterization gradient." Advances in neural information processing systems. 2016.