FINEST: Stabilizing Recommendations by Rank-Preserving Fine-Tuning
Not on the map. Why?
Abstract (arXiv)
Modern recommender systems may output considerably different recommendations due to small perturbations in the training data. Changes in the data from a single user will alter the recommendations as well as the recommendations of other users. In applications like healthcare, housing, and finance, this sensitivity can have adverse effects on user experience. We propose a method to stabilize a given recommender system against such perturbations. This is a challenging task due to (1) the lack of a ``reference'' rank list that can be used to anchor the outputs; and (2) the computational challenges in ensuring the stability of rank lists with respect to all possible perturbations of training data. Our method, FINEST, overcomes these challenges by obtaining reference rank lists from a given recommendation model and then fine-tuning the model under simulated perturbation scenarios with rank-preserving regularization on sampled items. Our experiments on real-world datasets demonstrate that FINEST can ensure that recommender models output stable recommendations under a wide range of different perturbations without compromising next-item prediction accuracy.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 48 of its 103 references that are in the map. First posted on arXiv 5 Feb 2024.
Builds on
48 of 103 references.
References · 48
- 2017Attention Is All You Needcited by 59,932
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,259
- 2018Self-Attentive Sequential Recommendationcited by 1,169
- 2019BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformercited by 952
- 2020Estimating Training Data Influence by Tracing Gradient Descentcited by 565
- 2020Underspecification Presents Challenges for Credibility in Modern Machine Learningcited by 488
- 2019Predicting Dynamic Embedding Trajectory in Temporal Interaction Networkscited by 443
- 2018Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Modelscited by 362
- 2019Multi-Task Deep Neural Networks for Natural Language Understandingcited by 349
- 2018Session-based Recommendation with Graph Neural Networkscited by 312
- 2017Recurrent Neural Networks with Top-k Gains for Session-based Recommendationscited by 165
36 more
- 2019Predictive Multiplicity in Classificationcited by 162
- 2020Certified Robustness to Label-Flipping Attacks via Randomized Smoothingcited by 103
- 2022A Survey on the Fairness of Recommender Systemscited by 102
- 2019A Survey on Session-based Recommender Systemscited by 97
- 2021Global Context Enhanced Graph Neural Networks for Session-based Recommendationcited by 81
- 2020Interactive Path Reasoning on Graph for Conversational Recommendationcited by 72
- 2018Ranking Distillation: Learning Compact Ranking Models With High Performance for Recommender Systemcited by 60
- 2019Robustness Certificates for Sparse Adversarial Attacks by Randomized Ablationcited by 59
- 2018RepeatNet: A Repeat Aware Neural Recommendation Machine for Session-based Recommendationcited by 51
- 2018Adversarial Personalized Ranking for Recommendationcited by 48
- 2022Predictive Multiplicity in Probabilistic Classificationcited by 44
- 2021Self-Supervised Graph Co-Training for Session-based Recommendationcited by 43
- 2020A Survey on Trajectory Data Management, Analytics, and Learningcited by 41
- 2022A Survey on Trustworthy Recommender Systemscited by 40
- 2021Leave-one-out Unfairnesscited by 36
- 2021Black-Box Attacks on Sequential Recommenders via Data-Free Model Extractioncited by 34
- 2021Dynamic Graph Collaborative Filteringcited by 33
- 2019Adversarial Attack Generation Empowered by Min-Max Optimizationcited by 32
- 2019Measuring the Business Value of Recommender Systemscited by 32
- 2023Linear Recurrent Units for Sequential Recommendationcited by 32
- 2020Revisiting Adversarially Learned Injection Attacks Against Recommender Systemscited by 30
- 2018Adversarial Training Towards Robust Multimedia Recommender Systemcited by 27
- 2020Practical Data Poisoning Attack against Next-Item Recommendationcited by 27
- 2018Adversarial Sampling and Training for Semi-Supervised Information Retrievalcited by 23
- 2022Rank List Sensitivity of Recommender Systems to Interaction Perturbationscited by 20
- 2020Incorporating User Micro-behaviors and Item Knowledge into Multi-task Learning for Session-based Recommendationcited by 18
- 2021Graph-Enhanced Multi-Task Learning of Multi-Level Transition Dynamics for Session-based Recommendationcited by 18
- 2021Churn Reduction via Distillationcited by 17
- 2022Untargeted Attack against Federated Recommendation Systems via Poisonous Item Embeddings and the Defensecited by 16
- 2022Defending Substitution-Based Profile Pollution Attacks on Sequential Recommenderscited by 11
- 2022M2TRec: Metadata-aware Multi-task Transformer for Large-scale and Cold-start free Session-based Recommendationscited by 9
- 2022Reducing Model Jitter: Stable Re-training of Semantic Parsers in Production Environmentscited by 6
- 2022Implicit Session Contexts for Next-Item Recommendationscited by 5
- 2021IACN: Influence-aware and Attention-based Co-evolutionary Network for Recommendationcited by 3
- 2023Towards More Robust and Accurate Sequential Recommendation with Cascade-guided Adversarial Trainingcited by 2
- 2023Investigating the Robustness of Sequential Recommender Systems Against Training Data Perturbationscited by 2
Built on by
Citing papers per year: 2024: 1; 2025: 1.
2 papers.