The Yahoo! Music Dataset and KDD-Cup '11

By Gideon Dror, Noam Koenigstein, Yehuda Koren, Markus Weimer. 2012

326 citations recorded by the source corpus.

How researchers see this paper

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

In Evaluation & Benchmarks

Citations in the corpus

Maximum inner-product search using cone trees (2012, methodology): “We use the following collaborative filtering datasets: MovieLens [17], Netflix [3] and the Yahoo! Music [12] datasets.”
Geometric Matrix Completion with Recurrent Multi-Graph Neural Networks (2017, methodology): “We closely followed the experimental setup of (Rao et al., 2015), using ﬁve standard datasets: Synthetic dataset from (Kalofolias et al., 2014), MovieLens (Miller et al., 2003), Flixster (Jamali & Ester, 2010), Douban (Ma et al., 2011), and YahooMusic (Dror et al., 2012).”
A Comprehensive Linear Speedup Analysis for Asynchronous Stochastic Parallel Optimization from Zeroth-Order to First-Order (2016, methodology): “2 Asynchronous Parallel Model Combination for Yahoo Music Recommendation Competition In KDD-Cup 2011, teams were challenged to predict user ratings in music given the Yahoo! Music data set [8].”
Fast and Robust Parallel SGD Matrix Factorization (2015, background): “Among those, SGD has gained much attention, as the winners of two extensive competitions, KDDCup 2011 [3] and Netﬂix comptition [1], used SGD-based matrix-factorization algorithms [1, 25].”
A fast parallel SGD for matrix factorization in shared memory systems (2013, background): “Categories and Subject Descriptors G.4 [Mathematics of Computing]: Mathematical Software—Parallel and vector implementations Keywords Recommender system, Matrix factorization, Stochastic gradient descent, Parallel computing, Shared memory algorithm”
Improve collaborative filtering through bordered block diagonal form matrices (2013, methodology): “The Yahoo Music dataset [8] is from KDD Cup 2011, and its ratings range from 1 to 100.”
The MovieLens Datasets: History and Context (2016, background): “Their “C15” dataset, released for the KDD Cup [Dror et al. 2012] provides user ratings on songs, albums, artists, and music genres on a 101 point (0-100) rating scale (though 98% of the ratings are multiples of 10).”
Fast Exact Max-Kernel Search (2012, methodology): “Extensions for approximate max-kernel search are also presented.”
A Fast Parallel Stochastic Gradient Method for Matrix Factorization in Shared Memory Systems (2015, background): “As demonstrated in KDD Cup 2011 [Dror et al. 2012] and Netflix competition [Bell and Koren 2007], a collaborative filter using latent factors has been considered as one of the best models for recommender systems.”
Scalable Recommendation with Hierarchical Poisson Factorization (2015, methodology): “When modeling implicit feedback data sets, researchers have proposed merging factorization techniques with neighborhood models [22], weighting techniques to adjust the relative importance of positive examples [17], and sampling-based approaches to create informative negative examples [8, 7, 29].”
Collaborative Filtering with Graph Information: Consistency and Scalable Methods (2015, unknown): “Dataset # users # items # ratings # links rank used Flixster ([11]) 147,612 48,794 8,196,077 2,538,746 10 Douban ([14]) 129,490 58,541 16,830,839 1,711,802 10 YahooMusic ([8]) 249,012 296,111 55,749,965 57,248,136 20”
One-class collaborative filtering with random graphs (2013, methodology): “This approach was effectively utilized by many of the leading solutions in the KDD-Cup’11 competition [4].”
Efficient retrieval of recommendations in a matrix factorization framework (2012, methodology): “Music respectively, because these values showed a good balance be-tween performance and speedup.”
NOMAD: Nonlocking, stOchastic Multi-machine algorithm for Asynchronous and Decentralized matrix completion (2013, unknown): “Name Rows Columns Non-zeros Netflix [5] 2,649,429 17,770 99,072,112 Yahoo! Music [10] 1,999,990 624,961 252,800,275 Hugewiki [2] 50,082,603 39,780 2,736,496,604”
Scalable Recommendation with Poisson Factorization (2013, methodology): “When modeling implicit feedback data sets, researchers have proposed merging factorization techniques with neighborhood models [20], weighting techniques to adjust the relative importance of positive examples [15], and samplingbased approaches to create informative negative examples [7, 6, 27].”
To Index or Not to Index: Optimizing Exact Maximum Inner Product Search (2017, unknown): “Dataset # users # items # ratings Netflix Prize [4] (Netflix) 480,189 17,770 100,480,507 Yahoo Music KDD [10] (KDD) 1,000,990 624,961 252,810,175 Yahoo Music R2 [39] (R2) 1,823,179 136,736 699,640,226 GloVe-Twitter [26] 100,000 1,093,514 –”
Partially Observable Markov Decision Process for Recommender Systems (2016, background): “1 [4], by the ofﬁcial KDDCUP (cid:48) 11 competition report [5].”