Performance of recommender algorithms on top-n recommendation tasks

By Paolo Cremonesi, Yehuda Koren, Roberto Turrin. 2010

1,543 citations recorded by the source corpus.

Subfields: Information Systems, Management Science and Operations Research, Marketing, Information Retrieval & Search

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Systems

In Management Science and Operations Research

In Marketing

In Information Retrieval & Search

Citations in the corpus

Quantization based Fast Inner Product Search (2015, background): “For example, a recommendation system needs to retrieve the most relevant items to a user from an inventory of millions of items, whose relevance is commonly represented as inner products [7].”
Asymmetric LSH (ALSH) for Sublinear Time Maximum Inner Product Search (MIPS) (2014, methodology): “Given the user-item ratings matrix R, we follow the standard PureSVD procedure [4] to generate user and item latent vectors.”
Evaluating Prerequisite Qualities for Learning End-to-End Dialog Systems (2015, background): “This is standard in recommendation, see e.g. Cremonesi et al. (2010).”
On the Effectiveness of Linear Models for One-Class Collaborative Filtering (2016, methodology): “• PureSVD of (Cremonesi, Koren, and Turrin 2010).”
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].”
How algorithmic confounding in recommendation systems increases homogeneity and decreases utility (2017, background): “Low errors for predicting ratings, however, do not always correspond to high accuracy in rank-based evaluation [14, 54] (e.”
Evaluation of recommendations: rating-prediction and ranking (2013, methodology): “Whilethere are several state-of-the-art approaches, e.g., [6, 9, 10, 16, 19], that achieve a lower RMSE on observed ratings than MF-RMSE does, note that their test performances on all (unrated) items is quite similar to each other, e.g., see [10, 4].”
Offline and online evaluation of news recommender systems at swissinfo.ch (2014, background): “While there are many different metrics to assess the performance of a recommender systems [5, 9], we focus on this metric because it can also be measured online and thus allows a comparison.”
Social Network and Tag Sources Based Augmenting Collaborative Recommender System (2015, background): “The recommender sorts the preferences in descending order, and ﬁnally chooses the top n items [18] as the recommendations for the test user.”
entity2rec: Learning User-Item Relatedness from Knowledge Graphs for Top-N Item Recommendation (2017, methodology): “As baselines, we use state-of-the-art collaborative ltering algorithms based on Non-Negative Matrix Factorization (NMF) [17], on Singular Value Decomposition [15], ItemKNN with baselines [14] and the Most Popular Items recommendation strategy, which is known to be quite eective on MovieLens due to”
On Symmetric and Asymmetric LSHs for Inner Product Search (2014, methodology): “, 2005; Cremonesi et al. , 2010), in multi-class prediction ( Dean et al. , 2013; Jain et al. , 2009) and structural SVM ( Joachims , 2006; Joachims et al. , 2009) problems and in vision problems when scoring filters based on their activations ( Dean et al. , 2013) (see Shrivastava and Li , 2014a, f”
Comparing the Predictive Capability of Social and Interest Affinity for Recommendations (2014, background): “Albeit RMSE ability to gauge the performance for pervasive top-k recommendations is debated [10], it best fits our purpose to measure performance shifts across classes of items/users.”
Explicit factor models for explainable recommendation based on phrase-level sentiment analysis (2014, methodology): “However, rating prediction accuracy is not necessarily related to the performance in practical applications [4].”
Sequence-Aware Recommender Systems (2018, methodology): “nted Cumulative Gain (NDCG) and the F1 metric. Typically, most of the ranking metric are highly correlated when evaluated with a fixed-size split and their choice does not largely affect the outcomes [24,44]. For some application domains, such as context adaptation, we can use standard classificatio”
Clustering is Efficient for Approximate Maximum Inner Product Search (2015, methodology): “Given the user-item matrix Z , we follow the pureSVD procedure described in (Cremonesi et al., 2010) to generate user and movie vectors.”
LCARS: a location-content-aware recommender system (2013, methodology): “To evaluate the recommender models, we adopt the testing methodology and the measurement Recall@k applied in [7, 5, 13, 32].”
Incorporating Phrase-level Sentiment Analysis on Textual Reviews for Personalized Recommendation (2015, background): “…Subject Descriptors H.3.3 [Information Storage and Retrieval]: Information Filtering; I.2.7 [Artificial Intelligence]: Natural Language Processing; H.3.5 [Online Information Services]: Webbased services Keywords Personalized Recommendation; Collaborative Filtering; Sentiment Analysis; Text Mining”
TFMAP: optimizing MAP for top-n context-aware recommendation (2012, methodology): “For this reason, we adopt a conventional widelyused evaluation strategy [9, 17], in which we randomly select 1000 items that have no feedback as irrelevant ones for each user in the test set.”
Improved Asymmetric Locality Sensitive Hashing (ALSH) for Maximum Inner Product Search (MIPS) (2014, methodology): “As reviewed in our prior work [23], examples of applications of MIPS include recommender system [16, 5, 15], large-scale object detection with DPM [9, 7, 14, 14], structural SVM [7], and multi-class label prediction [21, 15, 25].”
A survey of active learning in collaborative filtering recommender systems (2016, methodology): “The RMSE metric over the training set can be computed by adopting leaveone-out methodology [71].”