IRGAN: A Minimax Game for Unifying Generative and Discriminative Information Retrieval Models

Abstract

This paper provides a unified account of two schools of thinking in information retrieval modelling: the generative retrieval focusing on predicting relevant documents given a query, and the discriminative retrieval focusing on predicting relevancy given a query-document pair. We propose a game theoretical minimax game to iteratively optimise both models. On one hand, the discriminative model, aiming to mine signals from labelled and unlabelled data, provides guidance to train the generative model towards fitting the underlying relevance distribution over documents given the query. On the other hand, the generative model, acting as an attacker to the current discriminative model, generates difficult examples for the discriminative model in an adversarial way by minimising its discrimination objective. With the competition between these two models, we show that the unified framework takes advantage of both schools of thinking: (i) the generative model learns to fit the relevance distribution over documents via the signals from the discriminative model, and (ii) the discriminative model is able to exploit the unlabelled data selected by the generative model to achieve a better estimation for document ranking. Our experimental results have demonstrated significant performance gains as much as 23.96% on Precision@5 and 15.50% on MAP over strong baselines in a variety of applications including web search, item recommendation, and question answering.

12 pages; appendix added

Cited by in corpus (45)

• Deep Learning based Recommender System: A Survey and New Perspectives
• Adversarial Personalized Ranking for Recommendation
• Collaborative Memory Network for Recommendation Systems
• Spectral Collaborative Filtering
• Reinforced Negative Sampling over Knowledge Graph for Recommendation
• Elliot: a Comprehensive and Rigorous Framework for Reproducible Recommender Systems Evaluation
• CoaCor: Code Annotation for Code Retrieval with Reinforcement Learning
• ESAM: Discriminative Domain Adaptation with Non-Displayed Items to Improve Long-Tail Performance
• Sequential Recommendation with Self-Attentive Multi-Adversarial Network
• Adversarial Sampling and Training for Semi-Supervised Information Retrieval
• Generating Reliable Friends via Adversarial Training to Improve Social Recommendation
• Learning Robust Recommenders through Cross-Model Agreement
• Enhancing VAEs for Collaborative Filtering: Flexible Priors & Gating Mechanisms
• On the Theories Behind Hard Negative Sampling for Recommendation
• Multiresolution Graph Attention Networks for Relevance Matching
• Positive-Unlabeled Learning with Adversarial Data Augmentation for Knowledge Graph Completion
• Graph Contrastive Learning with Generative Adversarial Network
• Quantum semi-supervised generative adversarial network for enhanced data classification
• Content-Based Weak Supervision for Ad-Hoc Re-Ranking
• CnGAN: Generative Adversarial Networks for Cross-network user preference generation for non-overlapped users
• Augmented Negative Sampling for Collaborative Filtering
• GAN-based Matrix Factorization for Recommender Systems
• Fast Variational AutoEncoder with Inverted Multi-Index for Collaborative Filtering
• Rethinking Collaborative Metric Learning: Toward an Efficient Alternative without Negative Sampling
• Generative Flow Network for Listwise Recommendation
• Neighborhood-based Hard Negative Mining for Sequential Recommendation
• Learning Explicit User Interest Boundary for Recommendation
• Variation Control and Evaluation for Generative SlateRecommendations
• Dimension Independent Mixup for Hard Negative Sample in Collaborative Filtering
• Regularized Adversarial Sampling and Deep Time-aware Attention for Click-Through Rate Prediction
• Unlocking the Hidden Treasures: Enhancing Recommendations with Unlabeled Data
• Improve Learning from Crowds via Generative Augmentation
• High-Precision Extraction of Emerging Concepts from Scientific Literature
• Stimulate the Potential of Robots via Competition
• Adversarial Collaborative Filtering for Free
• UniPoll: A Unified Social Media Poll Generation Framework via Multi-Objective Optimization
• Generative Adversarial User Privacy in Lossy Single-Server Information Retrieval
• Collaborative Diffusion Model for Recommender System
• An Evaluation Study of Generative Adversarial Networks for Collaborative Filtering
• MixDec Sampling: A Soft Link-based Sampling Method of Graph Neural Network for Recommendation
• ESANS: Effective and Semantic-Aware Negative Sampling for Large-Scale Retrieval Systems
• Graph-Based Model-Agnostic Data Subsampling for Recommendation Systems
• Evaluating and Addressing Fairness Across User Groups in Negative Sampling for Recommender Systems
• Scalable Personalised Item Ranking through Parametric Density Estimation
• Exploiting Preferences in Loss Functions for Sequential Recommendation via Weak Transitivity