Maximum-Margin Matrix Factorization

By Nati Srebro, Jason D. M. Rennie, Tommi S. Jaakkola. 2004

1,237 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 Optimization & Training

Citations in the corpus

Hogwild: A Lock-Free Approach to Parallelizing Stochastic Gradient Descent (2011, background): “…low latency and high throughput shared main memory (a processor in such a system can write and read the shared physical memory at over 12GB/s with latency in the tens of nanoseconds); and (2) high bandwidth off multiple disks (a thousand-dollar RAID can pump data into main memory at over 1GB/s).”
Explainable Recommendation: A Survey and New Perspectives (2018, background): “Recommendation explanation can come in different forms in practice (Tintarev and Masthoff [2015]).”
Geometric Matrix Completion with Recurrent Multi-Graph Neural Networks (2017, methodology): “The geometric matrix completion problem thus boils down to minimizing A solution is to use a factorized representation (Srebro et al., 2004; Koren et al., 2009; Ma et al., 2011; Yanez & Bach, 2012; Rao et al., 2015; Benzi et al., 2016) X = WH (cid:62) , where W , H are m × r and n × r matrices,…”
Exploring Generalization in Deep Learning (2017, methodology): “Also norms such as the trace norm and max norm are considered as sensible inductive biases in matrix factorization and are often more appropriate than parameter-counting measures such as the rank [27, 28].”
Stronger generalization bounds for deep nets via a compression approach (2018, methodology): “…them has proved diﬃcult —at least in a quantitative way that yields sample size upper bounds similar to classical analyses in simpler models such as SVMs Bartlett and Mendelson [2002], Evgeniou et al. [2000], Smola et al. [1998] or matrix factorization Fazel et al. [2001], Srebro et al. [2005].”
Optimization with Sparsity-Inducing Penalties (2011, background): “Using the trace norm as a convex surrogate for rank has many applications in control theory [48], matrix completion [1, 130], multi-task learning [109], or multi-label classification [4], where low-rank priors are adapted.”
A Deep Framework for Cross-Domain and Cross-System Recommendations (2018, methodology): “Maximum-Margin Matrix Factorization (MMMF) [Srebro et al. , 2005] learns a matrix ^ R to ﬁt the observed rating matrix R by minimizing a trace norm of R and maximizing the corresponding predictive margin.”
Mind the gaps: weighting the unknown in large-scale one-class collaborative filtering (2009, methodology): “where h(·) is the Smooth Hinge loss function in [17], λ is a parameter trading off the weighted hinge loss and the margin and ‖ · ‖Σ is the trace norm of a matrix.”
SSE-PT: Sequential Recommendation Via Personalized Transformer (2020, methodology): “There are many other regularization techniques, including parameter sharing [7], max-norm regularization [28], gradient clipping [24], etc.”
Weighted Nuclear Norm Minimization and Its Applications to Low Level Vision (2016, background): “The related investigations range from studies on optimization algorithms (Jain et al, 2013) to new regularization strategies for highly ill-posed matrix completion tasks (Srebro et al, 2004).”
A Trace-restricted Kronecker-Factored Approximation to Natural Gradient (2020, neutral): “, the sum of singular values (Srebro et al., 2005; Recht et al., 2010).”
A Unified View of Matrix Factorization Models (2008, background): “Logistic Fast-MMMF is a variant of Fast-MMMF which uses log-loss and its matching link instead of smoothed Hinge loss, following [47].”
xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systems (2018, methodology): “ product of two latent vectors, thus it does not require the co-occurrence of user and item in the training set. MF is the most popular model-based collaborative filtering method in the RS literature [17, 20, 30, 38]. [4, 28] extend MF to leveraging side information, in which both a linear model and”