The MovieLens Datasets: History and Context

By F. Maxwell Harper, J. Konstan, J. A. 2016

5,680 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Science Applications, Information Systems, Evaluation & Benchmarks

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

In Computer Science Applications

In Information Systems

In Evaluation & Benchmarks

Citations in the corpus

DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning (2021, positive): “In the rest of this section, we present experiments on the following real MTL datasets: MovieLens, Multi-MNIST, Multi-Fashion MNIST, and on a real-world, large-scale recommender system.”
Predictability and Surprise in Large Generative Models (2022, background): “specific algorithms [34], but are not typically studied with large scale generative models (yet).”
On the challenges of learning with inference networks on sparse, high-dimensional data (2017, methodology): “Datasets: We study two large user-item rating datasets: MovieLens-20M (ML-20M) [Harper and Konstan, 2015] and Netflix(4).”
Deep Interest Network for Click-Through Rate Prediction (2017, background): “MovieLens data[10] contains 138,493 users, 27,278 movies, 21 categories and 20,000,263 samples.”
Compositional Fairness Constraints for Graph Embeddings (2019, methodology): “Our second dataset is derived from the MovieLens-1M recommender system benchmark (Harper & Konstan, 2016).”
How algorithmic confounding in recommendation systems increases homogeneity and decreases utility (2017, methodology): “For example, MovieLens is a research website that uses collaborative ltering to make movie recommendations [45]; the researchers behind it have released several public datasets6 that are used prolically in evaluation recommendation systems [23].”
Column Networks for Collective Classification (2016, methodology): “We used the MovieLens Latest Dataset (Harper and Konstan 2016) which consists of 33,000 movies.”
Federated Collaborative Filtering for Privacy-Preserving Personalized Recommendation System (2019, methodology): “The dataset [25] used in this research work contains 1,000,209 anonymous ratings of 3952 movies made by 6040 users who joined MovieLens in the year of 2000.”
Inferring Latent Network from Cascade Data for Dynamic Social Recommendation (2016, background): “Detailed descriptions of the data file can be found in [9].”
Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey (2019, background): “MovieLens-10M 28 : The MovieLens dataset [90] is a dynamic user-tag interactions dataset.”
Transparent, Scrutable and Explainable User Models for Personalized Recommendation (2019, methodology): “We employ the MovieLens-20M (ML-20M) dataset, which has been extensively used in recommendation research [20].”
Six Human-Centered Artificial Intelligence Grand Challenges (2023, neutral): “There have been a few examples of long-lived infrastructure including: the MovieLens research recommender system which has operated since 1997, though access has only been through co-operation with the University of Minnesota (Harper & Konstan, 2016); various universities operated Smart Homes, thoug”
Beyond Parity: Fairness Objectives for Collaborative Filtering (2017, methodology): “We use the Movielens Million Dataset [10], which contains ratings (from 1 to 5) by 6,040 users of 3,883 movies.”
TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation (2023, methodology): “This is a processed dataset from MovieLens100K [12] , which comprises user ratings on movies and comprehensive textual descriptions of movies such as “title” and “director”.”
Towards Deep Conversational Recommendations (2018, methodology): “We use a similar model and pre-train it with MovieLens data [5].”
Neural Logic Networks (2019, methodology): “We use a subset in the area of Electronics, containing 1,689,188 ratings ranging from 1 to 5 from 192,403 users and 63,001 items, which is bigger and much more sparse than the ML-100k dataset.”