Learning Tree-based Deep Model for Recommender Systems

By Han Zhu, Pengye Zhang, Guozheng Li, Jie He, Han Li, Kun Gai. Knowledge Discovery and Data Mining, 2018

346 citations recorded by the source corpus.

Model-based methods for recommender systems have been studied extensively in recent years. In systems with large corpus, however, the calculation cost for the learnt model to predict all user-item preferences is tremendous, which makes full corpus retrieval extremely difficult. To overcome the calculation barriers, models such as matrix factorization resort to inner product form (i.e., model user-item preference as the inner product of user, item latent factors) and indexes to facilitate efficient approximate k-nearest neighbor searches. However, it still remains challenging to incorporate more expressive interaction forms between user and item features, e.g., interactions through deep neural networks, because of the calculation cost. In this paper, we focus on the problem of introducing arbitrary advanced models to recommender systems with large corpus. We propose a novel tree-based method which can provide logarithmic complexity w.r.t. corpus size even with more expressive models such as deep neural networks. Our main idea is to predict user interests from coarse to fine by traversing tree nodes in a top-down fashion and making decisions for each user-node pair. We also show that the tree structure can be jointly learnt towards better compatibility with users' interest distribution and hence facilitate both training and prediction. Experimental evaluations with two large-scale real-world datasets show that the proposed method significantly outperforms traditional methods. Online A/B test results in Taobao display advertising platform also demonstrate the effectiveness of the proposed method in production environments.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/39667784085b83191b58c361d60d518a9e9f373d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xiangnan He
2. Tat-seng Chua
3. Chang Zhou
4. Wenqiang Lei
5. Hongxia Yang
6. Xiang Wang
7. Ed H. Chi
8. Jie Tang

In Information Systems

1. Xiangnan He
2. Tat-seng Chua
3. Ed H. Chi
4. Zhe Zhao
5. Wenqiang Lei
6. Xiang Wang
7. Lichan Hong
8. Maarten de Rijke

Citations in the corpus

• Controllable Multi-Interest Framework for Recommendation (2020, methodology): “from Amazon [18, 39]. In our experiment, we use the Books category of the Amazon dataset. Each training sample is truncated at length 20. •Taobao3 collectsuserbehaviorsfromTaobao’srecommender systems [61]. In our experiment, we only use the click behaviors and sort the behaviors from one user by tim”
• Behavior sequence transformer for e-commerce recommendation in Alibaba (2019, background): “In Alibaba, the largest ecommerce platform in China, RSs have been the key engine for its Gross Merchandise Volume (GMV) and revenues, and various deep learning based recommendation methods have been deployed in rich e-commerce scenarios [1, 8, 10, 11, 14, 15, 17, 18].”
• Practice on Long Sequential User Behavior Modeling for Click-Through Rate Prediction (2019, methodology): “Taobao Dataset3 is a collection of user behaviors from Taobao’s recommender system [12].”
• Advances and Challenges in Conversational Recommender Systems: A Survey (2021, background): “, 2018a), tree-based deep model (TDM) (Zhu et al., 2018), and graph convolutional networks (GCNs) (Ying et al.”
• Sampling-bias-corrected neural modeling for large corpus item recommendations (2019, background): “Recommendation systems help users discover content of interest across many internet services, including video recommendations [12, 18], app suggestions [9], and online advertisement targeting [38].”