New neural network predicts user-item relationships in evolving graphs.
problem Link prediction in dynamic graphs for recommendation services.
method Proposes a new neural network approach to leverage temporal contextual information.
result Our approach produces better predictions in scenarios with changing user-item relationships.
Proposes MCCF to distinguish latent purchasing motivations in user-item interactions.
problem Difficulty in capturing fine-grained user preferences due to complex latent motivations.
method Introduces MCCF with decomposer and combiner modules to identify and recombine latent components.
result Significant performance gains and necessity of considering multiple components demonstrated.
PGRec improves recommendation by modeling user-item preferences as a graph and embedding it for better predictions.
problem Sparse user-item data in recommender systems.
method PGRec models user-item preferences as a PrefGraph, then uses deep learning and factorization to embed and predict user preferences.
result PGRec outperforms state-of-the-art methods by up to 3.2% in NDCG@10.
New methods improve recommendation accuracy for users and items with few ratings.
problem Skewed distribution and low ratings affect recommendation accuracy.
method Four matrix completion-based approaches: FARP, TMF, TMF + Dropout, IFWMF.
result Improved prediction accuracy for users and items with few ratings.
JODIE learns dynamic user-item embeddings from interactions, outperforming existing methods.
problem Modeling dynamic user-item interactions for accurate future predictions.
method JODIE uses coupled recurrent models with update, projection, and prediction components, and a novel t-Batch algorithm.
result JODIE outperforms state-of-the-art methods by up to 22.4% on future interaction and state change prediction tasks.
New method uses bandit feedback to better evaluate recommender systems.
problem Traditional offline evaluation of recommender systems is inaccurate.
method Exploits bandit feedback to estimate online performance.
result Bandit feedback provides more accurate offline evaluation.
Develops a two-level monotonic multistage recommender system for better user-specific prediction.
problem Leveraging user-item-stage dependencies in a monotonic chain of events for enhanced prediction accuracy.
method A multistage recommender system with a two-level monotonic property, using a large-margin classifier based on a nonnegative additive latent factor model.
result The proposed method outperforms existing methods in simulations and an article sharing dataset.
DBRec discovers latent groups to improve recommendation.
problem Sparse user-item interaction data in recommender systems.
method Simultaneously discovers latent user/item groups and interacts them with users/items for bridging preferences.
result DBRec outperforms state-of-the-art models on real datasets.
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, mod…
TransCF improves recommendation by modeling user-item relationships with translation vectors.
problem Triangle inequality violation in matrix factorization-based recommendation methods.
method TransCF uses translation vectors to model latent user-item relationships in implicit feedback.
result TransCF outperforms state-of-the-art methods by up to 17% in hit ratio.
GEMRank embeds users and items using co-occurrence relations for better collaborative filtering.
problem Lack of textual data for entity embedding in recommender systems.
method Uses profile co-occurrence for entity relations and factorization for embedding. Feeds embeddings into a neural network for predictions.
result Significantly outperforms baseline algorithms in various data sets.
Improved item recommendations for repeat interactions using sequence analysis.
problem Limited effectiveness of traditional recommender systems in handling repeated user-item interactions.
method Designed a recommender system that considers sequences of item interactions for each user.
result Empirically shown to give highly accurate predictions and increase sales by 5%.
SAIN integrates user-item feedback with content attributes for better recommendation.
problem Cold start problems in recommendation models due to sparse user-item interactions.
method SAIN uses a self-attention mechanism to capture feature interactions and an information integration layer to combine feedback and content information.
result SAIN outperforms state-of-the-art models by 2.13% on public datasets.
Improved neural model for social recommendation by integrating social and interest networks.
problem Data sparsity and lack of higher-order relationships in social recommendation.
method DiffNet++ models neural influence diffusion and interest diffusion in a unified framework using a multi-level attention network.
result Extensive experiments on real-world datasets show the effectiveness of DiffNet++.
IGMC learns inductive matrix completion without side info.
problem Inductive matrix completion without side information.
method Graph Neural Network (GNN) trained on 1-hop subgraphs of the rating matrix.
result Achieves competitive performance with state-of-the-art transductive baselines.
New model improves recommendation systems by analyzing user-item interactions.
problem Improving recommendation systems for better user-item interactions.
method Sliced Anti-symmetric Decomposition (SAD) model using tensor decomposition.
result SAD produces the most consistent personalized preferences compared to SOTA models.
Matrix factorization techniques have been widely used as a method for collaborative filtering for recommender systems. In recent times, different variants of deep learning algorithms have been explored in this setting to improve the task of making a personalized recommendation with user-item interaction data. The idea …
Hybrid Deep Embedding for aspect-level explanations in recommendations.
problem Challenges in personalization, dynamic explanations, and aspect-level granularity in recommendation systems.
method Proposes Hybrid Deep Embedding (HDE) to learn dynamic embeddings for user and item preferences, and aspect-level quality vectors.
result Demonstrates improved recommending performance and dynamic aspect-level explanations.
Expands small recommendation datasets to industrial scale.
problem Disconnection between academic and industrial data scales.
method Randomized fractal expansions using Kronecker Graph Theory.
result Generated synthetic data sets with 1.2B ratings, 2.2M users, and 855K items.
We develop a Bayesian Poisson matrix factorization model for forming recommendations from sparse user behavior data. These data are large user/item matrices where each user has provided feedback on only a small subset of items, either explicitly (e.g., through star ratings) or implicitly (e.g., through views or purchas…
A new algorithm MBMF improves recommendation accuracy and speed for sparse datasets.
problem Sparse and fluctuating predictions in recommender systems.
method MBMF uses magnitude constraints and Spherical coordinates to optimize faster than existing methods.
result MBMF outperforms existing algorithms in accuracy and speed on synthetic and real datasets.
Matrix factorization simplifies user-item co-occurrence analysis.
problem Understanding the meaning of low-dimensional matrices in matrix factorization.
method Showed matrix factorization equals calculating eigenvectors of co-occurrence matrices, using RMT insights.
result Low-dimension matrices represent a reduced noise user and item co-occurrence space.
A new linear GCN model improves recommendation performance for large graphs.
problem Training difficulties and over-smoothing in GCN-based CF models.
method Proposes a linear residual graph convolutional network (LRGCCF) to address training difficulties and over-smoothing issues.
result The proposed model yields better efficiency and effectiveness on real datasets.
New cold-start recommendation method for recommender systems.
problem Making accurate recommendations for users and items with limited data.
method Proposes a faster cold-start prediction formula in collective matrix factorization models.
result Improved cold-start recommendations with reliable predictions for new users, better than non-personalized recommendations.
Generates massive synthetic data sets for recommender systems.
problem Size gap between academic data sets and industrial production systems.
method Expands pre-existing public data sets using Kronecker Graph Theory.
result Preserves higher order statistical properties of user/item interactions.
The goal of a recommendation system is to predict the interest of a user in a given item by exploiting the existing set of ratings as well as certain user/item features. A standard approach to modeling this problem is Inductive Matrix Completion where the predicted rating is modeled as an inner product of the user and …
We consider matrix completion for recommender systems from the point of view of link prediction on graphs. Interaction data such as movie ratings can be represented by a bipartite user-item graph with labeled edges denoting observed ratings. Building on recent progress in deep learning on graph-structured data, we prop…
We present our solution to the job recommendation task for RecSys Challenge 2016. The main contribution of our work is to combine temporal learning with sequence modeling to capture complex user-item activity patterns to improve job recommendations. First, we propose a time-based ranking model applied to historical obs…
This paper tackles selection bias in recommender systems by considering the neighborhood effect.
problem Selection bias in recommender systems due to filtering and user selection.
method Formalizes neighborhood effect as interference problem, introduces treatment representation, and proposes ideal loss.
result Proposed methods achieve unbiased learning when both selection bias and neighborhood effect are present.
CGM combines SSL and LFM for better recommendation performance.
problem Label sparsity in user-item rating matrices limits LFM performance.
method Probabilistic chain graph model (CGM) integrating Bayesian network and Markov random field.
result CGM significantly outperforms state-of-the-art approaches in recommendation.
Collaborative filtering (CF) aims to predict users' ratings on items according to historical user-item preference data. In many real-world applications, preference data are usually sparse, which would make models overfit and fail to give accurate predictions. Recently, several research works show that by transferring k…
Unified neural framework for multi-relational recommender systems.
problem Accurately capturing users' fine-grained preferences from diverse feedback types.
method Multi-Relational Memory Network (MRMN) framework that models fine-grained user-item relations and discriminates between feedback types.
result The proposed MRMN model outperforms state-of-the-art algorithms in various recommender scenarios.
Model-based collaborative filtering analyzes user-item interactions to infer latent factors that represent user preferences and item characteristics in order to predict future interactions. Most collaborative filtering algorithms assume that these latent factors are static, although it has been shown that user preferen…
Integrates contextual constraints into embedding models for better recommendation quality.
problem Contextual constraints lead to incomplete or low-quality recommendations when applied independently.
method Merges constraint application and retrieval into one operation in the embedding space.
result Significant improvements in predictive performance compared to context-aware and standard models.
Proposes a graph neural network for personalized news recommendation.
problem Data sparsity in news recommendation systems.
method Heterogeneous graph model + Graph Neural Networks + LSTM attention mechanism.
result Significantly outperforms state-of-the-art methods on news recommendation datasets.
The paper explores using set-level ratings for better user-item preference prediction in recommender systems.
problem Capturing user preferences on individual items using set-level ratings.
method Developed collaborative filtering-based methods to model user behaviors in set-level ratings.
result Collaborative filtering-based models can recover and predict user preferences on individual items using set-level ratings.
This paper introduces new methods to improve 1-bit matrix completion by considering cluster effects.
problem Improving 1-bit matrix completion for clustered data.
method Group-Specific 1-bit Matrix Completion (GS1MC) and Cluster Developing Matrix Completion (CDMC).
result GS1MC and CDMC outperform existing methods in synthetic and real-world data.
We tackle the problem of collaborative filtering (CF) with side information, through the lens of Gaussian Process (GP) regression. Driven by the idea of using the kernel to explicitly model user-item similarities, we formulate the GP in a way that allows the incorporation of low-rank matrix factorisation, arriving at o…
Matrix completion models are among the most common formulations of recommender systems. Recent works have showed a boost of performance of these techniques when introducing the pairwise relationships between users/items in the form of graphs, and imposing smoothness priors on these graphs. However, such techniques do n…
RNNs excel at both short-term and long-term user interaction prediction.
problem Modeling user preferences over short and long time horizons.
method Evaluating RNN-based models on both short-term and long-term recommendation tasks.
result RNNs can predict immediate and distant user interactions.
The paper proposes a new method for product recommendation that considers revenue contributions and user similarity.
problem High dimensionality and sparsity in user-item data, especially in terms of revenue contributions.
method The approach encodes revenue contributions in the user-item matrix and computes customer similarity using suitable distance measures.
result The method segments users based on revenue-based similarity and supports recommendations aligned with profitability objectives.
A new method speeds up ALS for recommender systems by subsampling key elements.
problem High computational cost of ALS for large-scale datasets.
method Core-elements subsampling method for efficient ALS approximation.
result Achieves similar accuracy with significantly reduced computational time.
KGAT uses knowledge graphs to improve recommendation accuracy and explainability.
problem Accurate, diverse, and explainable recommendations require side information and collaborative signals.
method KGAT models high-order relations in a knowledge graph by propagating embeddings and using attention mechanisms.
result KGAT significantly outperforms state-of-the-art methods on public benchmarks.
Predict artist efficiency in VFX shots using matrix completion.
problem Predicting artist efficiency in rendering VFX shots.
method Structured matrix factorization models for bounded entries.
result Effective models for predicting artist efficiency.
HGP tackles noisy connections and heterogeneity in social graphs for better social recommendation.
problem Challenges in social graphs, especially noisy connections and heterogeneity, hinder GNN-based social recommendation.
method HGP uses a tripartite graph to reduce complexity, personalized PageRank for node embeddings, and attention mechanism for integration.
result HGP outperforms baselines in AUC and F1-score metrics on a large-scale dataset.
Bayesian method predicts individual and crowd preferences from small data.
problem Difficult to predict preferences from limited personal data and noisy labels.
method Combines matrix factorization with Gaussian processes for scalable inference.
result Method predicts preferences for new users and items not in training set.
This paper introduces our solution to the 2018 Duolingo Shared Task on Second Language Acquisition Modeling (SLAM). We used deep factorization machines, a wide and deep learning model of pairwise relationships between users, items, skills, and other entities considered. Our solution (AUC 0.815) hopefully managed to bea…
A method learns matrix factorization from diverse matrices and applies the knowledge to unseen matrices.
problem Matrix factorization without shared rows or columns.
method Neural network meta-learned to minimize expected imputation error using MAP estimation.
result The method can impute missing values from unseen matrices efficiently.