One-Class Collaborative Filtering
By Rong Pan, Yunhong Zhou, Bin Cao, Nathan Nan Liu, Rajan M. Lukose, Martin Scholz, Qiang Yang. 2008
1,084 citations recorded by the source corpus.
Many applications of collaborative filtering (CF), such as news item recommendation and bookmark recommendation, are most naturally thought of as one-class collaborative filtering (OCCF) problems. In these problems, the training data usually consist simply of binary data reflecting a user's action or inaction, such as page visitation in the case of news item recommendation or webpage bookmarking in the bookmarking scenario. Usually this kind of data are extremely sparse (a small fraction are positive examples), therefore ambiguity arises in the interpretation of the non-positive examples. Negative examples and unlabeled positive examples are mixed together and we are typically unable to distinguish them. For example, we cannot really attribute a user not bookmarking a page to a lack of interest or lack of awareness of the page. Previous research addressing this one-class problem only considered it as a classification task. In this paper, we consider the one-class problem under the CF setting. We propose two frameworks to tackle OCCF. One is based on weighted low rank approximation; the other is based on negative example sampling. The experimental results show that our approaches significantly outperform the baselines.
Subfields: Artificial Intelligence, Information Systems, Recommender Systems
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
- Xiangnan He
- Tat-seng Chua
- Julian McAuley
- Charu C. Aggarwal
- Qiang Yang
- Philip S. Yu
- Yongfeng Zhang
- Xiang Wang
In Information Systems
- Xiangnan He
- Tat-seng Chua
- Xiang Wang
- Julian McAuley
- Yongfeng Zhang
- Charu C. Aggarwal
- Xing Xie
- Xu Chen
In Recommender Systems
Citations in the corpus
- On Sampling Strategies for Neural Network-based Collaborative Filtering (2017, background): “This is usually referred as “implicit feedback” [13, 23, 26].”
- Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering (2016, methodology): “We build scalable models to capture temporal dynamics in order to make better recommendations for the classical OneClass Collaborative Filtering setting [27], where only the implicit (or ‘positive’) feedback of users (i.”
- Variational Autoencoders for Collaborative Filtering (2018, methodology): “tasets, including two recently proposed neural-network approaches. 2 METHOD We use u ∈{1,...,U}to index users and i ∈{1,...,I}to index items. In this work, we consider learning with implicit feedback [19, 34]. The user-by-item interaction matrix is the click1 matrix X ∈NU×I. The lower case xu = [xu1”
- VBPR: Visual Bayesian Personalized Ranking from Implicit Feedback (2015, background): “Hu, Ko-ren, and Volinsky (2008) associate different ‘confidence lev-els’ to positive and non-observed feedback and then factor-ize the resulting weighted matrix, while Pan et al. (2008) sample non-observed feedback as negative instances and factorize a similar weighted matrix.”
- On the Effectiveness of Linear Models for One-Class Collaborative Filtering (2016, methodology): “Implicit feedback scenarios are also referred to as one-class collaborative filtering (OC-CF) problems (Pan et al. 2008; Pan and Scholz 2009); we use the terms interchangeably.”
- Gaussian Ranking by Matrix Factorization (2015, background): “There has been a large body of work on training collaborative filtering approaches as to maximize various ranking metrics, see eg [5, 8, 17, 2, 18, 10, 20, 21, 14, 13, 12].”
- BPR: Bayesian Personalized Ranking from Implicit Feedback (2009, methodology): “Machine learning approaches for item recommenders [5, 10] typically create the training data from S by giving pairs (u, i) ∈ S a positive class label and all other combinations in (U × I) \ S a negative one (see Figure 1).”
- Joint Deep Modeling of Users and Items Using Reviews for Recommendation (2017, methodology): “The CTR model solves the one-class collaborative filtering problem [23] by using two different values for the precision parameter c of a Gaussian distribution.”
- Partially Observable Markov Decision Process for Recommender Systems (2016, background): “feedback or the users miss the recommendations [20].”
- Collaborative Filtering-Based Recommendation of Online Social Voting (2017, methodology): “…these challenges, we develop a set of novel RS models, including matrix-factorization (MF)-based models and nearest-neighbor (NN)-based models, to learn user-voting interests by simultaneously mining information on user-voting participation, user–user friendship, and usergroup affliction.”
- Calibrated recommendations (2018, methodology): “As to generate the baseline-recommendations, we learned a weighted 50-dimensional matrix-factorization model on these training data following the usual approaches described in [14, 18, 23], where we tuned the hyper-parameters, i.”
- One-class classification: taxonomy of study and review of techniques (2013, neutral): “Pan et al .”
- Content-Aware Point of Interest Recommendation on Location-Based Social Networks (2015, background): “(1), following the standard strategy of solving one-class CF problems (Pan and Scholz 2009; Pan et al. 2008), we sample 10% of unobserved check-ins from the training matrix, deem them as the check-in frequency of 0 and set their corresponding Wi j to 1.”
- Improving pairwise learning for item recommendation from implicit feedback (2014, background): “[12] investigate item recommendation from implicit feedback and propose to cast the oneclass problem into a two class problem by imputing all nonobserved values with 0 and to apply regression.”
- Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding (2018, methodology): “, collaborative ltering [24], matrix factorization [15, 22], and top-N recommendation [9][19], are not suitable for capturing sequential patterns because they do not model the order of actions.”
- Item recommendation on monotonic behavior chains (2018, methodology): “In practice, cui can be implemented through sampling techniques during training [5, 17].”
- Collaborative matrix factorization with multiple similarities for predicting drug-target interactions (2013, methodology): “WLRA with Tikhonov regularization over A and B is also equivalent to one formalization for recommender systems, called one-class collaborative filtering, where an alternating least squares algorithm was presented for estimating A and B [20, 19, 15].”
- DELF: A Dual-Embedding based Deep Latent Factor Model for Recommendation (2018, background): “Nevertheless, since not all the un-observed entries are true negative instances, the solution may hinder useful information from the limited observed interactions and consequently degrade the quality of the learned user and item embeddings [ Pan et al. , 2008 ] .”
- CLiMF: learning to maximize reciprocal rank with collaborative less-is-more filtering (2012, methodology): “In addition, our work directly takes into account an evaluation metric, MRR, when developing the recommendation model, which is also substantially different from the work of [12, 19].”