Rating Augmentation with Generative Adversarial Networks towards Accurate Collaborative Filtering

Chae, Dong-Kyu; Kang, Jin-Soo; Kim, Sang–Wook; Choi, Joon Weon

doi:10.1145/3308558.3313413

Cited by 58 publications

(32 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Through the interplay between these two components, the model reaches the Nash equilibrium where G has learned to mimic the ground-truth data distribution, e.g., a profile of a particular user. In the present survey, we identified different application for GAN-based RS that include, improving negative sampling step in learning-to-rank objective function [39,126], fitting the generator to predict missing ratings by leveraging both temporal [8,147] and sideinformation [18,125], or augmenting training dataset [17,37].…”

Section: Cf Models Since Early Yearsmentioning

confidence: 99%

“…GANs have been shown powerful in generating relevant recommendations -in particular, using the CF approach -and capable of successively competing with state-of-the-art models in the field of RS. We have identified the following reasons for the potential of GANs in RS: (i) they are able to generalize well and learn unknown user preference distributions and thus be able to model user preference in complex settings (e.g., IRGAN [125] and CFGAN [18]); (ii) they are capable of generating more negative samples than random samples in pairwise learning tasks (e.g., APL [108], DASO [39]) and (iii) they can be used for data augmentation (e.g., AugCF [127] and RAGAN [17]).…”

Section: Collaborative Recommendationmentioning

confidence: 99%

See 1 more Smart Citation

Adversarial Machine Learning in Recommender Systems (AML-RecSys)

Deldjoo

Noia

Merra

2020

Proceedings of the 13th International Conference on Web Search and Data Mining

View full text Add to dashboard Cite

Latent-factor models (LFM) based on collaborative filtering (CF), such as matrix factorization (MF) and deep CF methods, are widely used in modern recommender systems (RS) due to their excellent performance and recommendation accuracy. Notwithstanding their great success, in recent years, it has been shown that these methods are vulnerable to adversarial examples, i.e., subtle but non-random perturbations designed to force recommendation models to produce erroneous outputs. The main reason for this behavior is that user interaction data used for training of LFM can be contaminated by malicious activities or users' misoperation that can induce an unpredictable amount of natural noise and harm recommendation outcomes. On the other side, it has been shown that these systems, conceived originally to attack machine learning applications, can be successfully adopted to strengthen their robustness against attacks as well as to train more precise recommendation engines.In this respect, the goal of this survey is two-fold: (i) to present recent advances on AML-RS for the security of RS (i.e., attacking and defense recommendation models), (ii) to show another successful application of AML in generative adversarial networks (GANs), which use the core concept of learning in AML (i.e., the min-max game) for generative applications.In this survey, we provide an exhaustive literature review of 60 articles published in major RS and ML journals and conferences. This review serves as a reference for the RS community, working on the security of RS and recommendation models leveraging generative models to improve their quality.

show abstract

Section: Cf Models Since Early Yearsmentioning

confidence: 99%

Section: Collaborative Recommendationmentioning

confidence: 99%

Adversarial Machine Learning in Recommender Systems (AML-RecSys)

Deldjoo

Noia

Merra

2020

Proceedings of the 13th International Conference on Web Search and Data Mining

View full text Add to dashboard Cite

show abstract

“…To address this issue, IRGAN [22], the first to introduce GAN into recommendation problems, employs the policy gradient strategy to estimate the gradients, while other models adopt NNs (e.g. multi-layer perceptrons (MLP)) as fake rating generators to enable stochastic gradient descent (SGD) [2]. GAIN [25] generatively imputes the missing ratings in the user-item matrix as an data augmentation approach by a fully-connected neural network.…”

Section: Adversarial Learning In Recommendationmentioning

confidence: 99%

“…Ciao 2 The CiaoDVD dataset is collected from the Ciao website dvd.ciao.co.uk in December, 2013. In this dataset, some users give repetitive ratings to the same item at different timestamps.…”

Section: Datasetsmentioning

confidence: 99%

Exploring Missing Interactions: A Convolutional Generative Adversarial Network for Collaborative Filtering

Yuan

Yao

Benatallah

2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

Adversarial examples can be detrimental to a recommender,leading to a surging enthusiasm for applying adversarial learning to improve recommendation performance, e.g. raising model robustness, alleviating data sparsity, generating initial profiles for cold-start users or items, etc. Most existing adversarial example generation methods fall within three categories: attacking the user-item interactions or auxiliary contents, adding perturbations in latent space, sampling the latent space according to certain distribution. In this work, we focus on the semantic-rich user-item interactions in a recommender system and propose a novel generative adversarial network (GAN) named Convolutional Generative Collaborative Filtering (Conv-GCF). We develop an effective perturbation mechanism (adversarial noise layer) for convolutional neural networks (CNN), based on which we design a generator with residual blocks to synthesize user-item interactions. We empirically demonstrate that on Conv-GCF, the adversarial noise layer is superior to the conventional noise-adding approach. Moreover, we propose two types of discriminators: one using Bayes Personalized Ranking (BPR) and the other with binary classification. On four public datasets, we show that our approach achieves the state-of-the-art top-n recommendation performance among competitive baselines.

show abstract

“…However, this does not well-address the "biased learning" problems. Notably, recent studies attempt to incorporate adversarial training as a new direction to alleviate the cold-start problems in recommendation, which performs data augmentation by imposing adversarial perturbations [1,2,20].…”

Section: Introductionmentioning

confidence: 99%

Comparative Study of Adversarial Training Methods for Cold-Start Recommendation

Meng

et al. 2021

Proceedings of the 1st International Workshop on Adversarial Learning for Multimedia

View full text Add to dashboard Cite

Adversarial training in recommendation is originated to improve the robustness of recommenders to attack signals and has recently shown promising results to alleviate cold-start recommendation. However, existing methods usually should make a trade-off between model robustness and performance, and the underlying reasons why using adversarial samples for training works has not been sufficiently verified. To address this issue, this paper identifies the key components of existing adversarial training methods and presents a taxonomy that defines these methods using three levels of components for perturbation generation, perturbation incorporation, and model optimization. Based on this taxonomy, different variants of existing methods are created, and a comparative study is conducted to verify the influence of each component in cold-start recommendation. Experimental results on two benchmarking datasets show that existing state-of-the-art algorithms can be further improved by a proper pairing of the key components as listed in the taxonomy. Moreover, using case studies and visualization, the influence of the content information of items on cold-start recommendation has been analyzed, and the explanations for the working mechanism of different components as proposed in the taxonomy have been offered. These verify the effectiveness of the proposed taxonomy as a design paradigm for adversarial training.

show abstract

Rating Augmentation with Generative Adversarial Networks towards Accurate Collaborative Filtering

Cited by 58 publications

References 24 publications

Adversarial Machine Learning in Recommender Systems (AML-RecSys)

Adversarial Machine Learning in Recommender Systems (AML-RecSys)

Exploring Missing Interactions: A Convolutional Generative Adversarial Network for Collaborative Filtering

Comparative Study of Adversarial Training Methods for Cold-Start Recommendation

Contact Info

Product

Resources

About