Learning Tree-based Deep Model for Recommender Systems

Abstract

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.

Accepted by KDD 2018

References in corpus (9)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• BPR: Bayesian Personalized Ranking from Implicit Feedback
• Real-Time Bidding with Multi-Agent Reinforcement Learning in Display Advertising
• Optimized Cost per Click in Taobao Display Advertising
• Collaborative Filtering with Recurrent Neural Networks
• Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction
• Product-based Neural Networks for User Response Prediction
• Deep Interest Network for Click-Through Rate Prediction
• Image Matters: Visually modeling user behaviors using Advanced Model Server

Cited by in corpus (63)

• Practice on Long Sequential User Behavior Modeling for Click-Through Rate Prediction
• DGCN: Diversified Recommendation with Graph Convolutional Networks
• Lifelong Sequential Modeling with Personalized Memorization for User Response Prediction
• STP-UDGAT: Spatial-Temporal-Preference User Dimensional Graph Attention Network for Next POI Recommendation
• Disentangling Long and Short-Term Interests for Recommendation
• User Behavior Retrieval for Click-Through Rate Prediction
• Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems
• How to Index Item IDs for Recommendation Foundation Models
• Graph Contrastive Learning with Generative Adversarial Network
• Sequential Recommendation with Dual Side Neighbor-based Collaborative Relation Modeling
• Behavior Sequence Transformer for E-commerce Recommendation in Alibaba
• Joint Optimization of Tree-based Index and Deep Model for Recommender Systems
• Improving Implicit Feedback-Based Recommendation through Multi-Behavior Alignment
• DaisyRec 2.0: Benchmarking Recommendation for Rigorous Evaluation
• Joint Learning of Deep Retrieval Model and Product Quantization based Embedding Index
• Economic Recommender Systems -- A Systematic Review
• Revisiting Neural Retrieval on Accelerators
• CheatAgent: Attacking LLM-Empowered Recommender Systems via LLM Agent
• Query Rewriting via Cycle-Consistent Translation for E-Commerce Search
• SDM: Sequential Deep Matching Model for Online Large-scale Recommender System
• Controllable Multi-Interest Framework for Recommendation
• EdgeRec: Recommender System on Edge in Mobile Taobao
• We Know What You Want: An Advertising Strategy Recommender System for Online Advertising
• Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems
• Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning
• On component interactions in two-stage recommender systems
• COLD: Towards the Next Generation of Pre-Ranking System
• Personalized Embedding-based e-Commerce Recommendations at eBay
• Entire Space Multi-Task Modeling via Post-Click Behavior Decomposition for Conversion Rate Prediction
• Cooperative Retriever and Ranker in Deep Recommenders
• ARGO: Modeling Heterogeneity in E-commerce Recommendation
• Image Matters: Visually modeling user behaviors using Advanced Model Server
• Dynamic Memory based Attention Network for Sequential Recommendation
• CROLoss: Towards a Customizable Loss for Retrieval Models in Recommender Systems
• Towards Personalized and Semantic Retrieval: An End-to-End Solution for E-commerce Search via Embedding Learning
• Fuzzy Norm-Explicit Product Quantization for Recommender Systems
• A Framework for Elastic Adaptation of User Multiple Intents in Sequential Recommendation
• Constrained Auto-Regressive Decoding Constrains Generative Retrieval
• Learning to Structure Long-term Dependence for Sequential Recommendation
• DLF: Enhancing Explicit-Implicit Interaction via Dynamic Low-Order-Aware Fusion for CTR Prediction
• Criterion-based Heterogeneous Collaborative Filtering for Multi-behavior Implicit Recommendation
• Improving Accuracy and Diversity in Matching of Recommendation with Diversified Preference Network
• Deep Retrieval: Learning A Retrievable Structure for Large-Scale Recommendations
• Context-aware Deep Model for Entity Recommendation in Search Engine at Alibaba
• Spatial-Temporal Deep Intention Destination Networks for Online Travel Planning
• Retrieval with Learned Similarities
• CoSam: An Efficient Collaborative Adaptive Sampler for Recommendation
• Show Me the Whole World: Towards Entire Item Space Exploration for Interactive Personalized Recommendations
• Scene Understanding in Pick-and-Place Tasks: Analyzing Transformations Between Initial and Final Scenes
• One Backward from Ten Forward, Subsampling for Large-Scale Deep Learning
• A Deep Recurrent Survival Model for Unbiased Ranking
• Relevance Proximity Graphs for Fast Relevance Retrieval
• Context-aware Tree-based Deep Model for Recommender Systems
• Itinerary-aware Personalized Deep Matching at Fliggy
• Sparse-Interest Network for Sequential Recommendation
• MIC: Model-agnostic Integrated Cross-channel Recommenders
• Origin-Aware Next Destination Recommendation with Personalized Preference Attention
• Semantic-enhanced Modality-asymmetric Retrieval for Online E-commerce Search
• $C^{3}DRec$: Cloud-Client Cooperative Deep Learning for Temporal Recommendation in the Post-GDPR Era
• LHRM: A LBS based Heterogeneous Relations Model for User Cold Start Recommendation in Online Travel Platform
• MAC: A Conversion Rate Prediction Benchmark Featuring Labels Under Multiple Attribution Mechanisms
• Deep Personalized Re-targeting
• Truncation-Free Matching System for Display Advertising at Alibaba