Embarrassingly Shallow Autoencoders for Sparse Data

Abstract

Combining simple elements from the literature, we define a linear model that is geared toward sparse data, in particular implicit feedback data for recommender systems. We show that its training objective has a closed-form solution, and discuss the resulting conceptual insights. Surprisingly, this simple model achieves better ranking accuracy than various state-of-the-art collaborative-filtering approaches, including deep non-linear models, on most of the publicly available data-sets used in our experiments.

In the proceedings of the Web Conference (WWW) 2019 (7 pages)

References in corpus (2)

• Iterative Thresholding Algorithm for Sparse Inverse Covariance Estimation
• Collaborative Filtering via High-Dimensional Regression

Cited by in corpus (18)

• Analyzing Item Popularity Bias of Music Recommender Systems: Are Different Genders Equally Affected?
• Top-N Recommendation Algorithms: A Quest for the State-of-the-Art
• On Sampling Top-K Recommendation Evaluation
• Reenvisioning Collaborative Filtering vs Matrix Factorization
• Towards a Better Understanding of Linear Models for Recommendation
• S-Walk: Accurate and Scalable Session-based Recommendationwith Random Walks
• Local Collaborative Autoencoders
• Collaborative Filtering via High-Dimensional Regression
• Do Offline Metrics Predict Online Performance in Recommender Systems?
• Revisiting Neighborhood-based Link Prediction for Collaborative Filtering
• Measuring the User Satisfaction in a Recommendation Interface with Multiple Carousels
• On Estimating Recommendation Evaluation Metrics under Sampling
• Density-Ratio Based Personalised Ranking from Implicit Feedback
• How Powerful is Graph Convolution for Recommendation?
• iALS++: Speeding up Matrix Factorization with Subspace Optimization
• NeuSE: A Neural Snapshot Ensemble Method for Collaborative Filtering
• I know why you like this movie: Interpretable Efficient Multimodal Recommender
• On the Regularization of Autoencoders