Sparse coding and NMF

By Julian Eggert, Edgar Körner. 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541), 2004

337 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/17885e2dfde8542d0821ff8ba67caf84c7e71ec8

How researchers see this paper

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

In Cognitive Neuroscience

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Energy Disaggregation via Discriminative Sparse Coding (2010, methodology): “To solve the optimization over Bi in step 2b, we use the multiplicative non-negative matrix fa ctorization update from [7].”
Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing (2019, methodology): “to outperform both a conventional deep neural network [26] and the iterative sparse NMF method [42].”
New Era for Robust Speech Recognition (2017, methodology): “An example of a wideband approach is that of [29], which combines ideas from [10, 28, 53].”
A General Flexible Framework for the Handling of Prior Information in Audio Source Separation (2012, methodology): “, to have a unitary (say ) norm, which can be handled using the gradient descent updates from [46] or the modified multiplicative updates from [47].”
Deep Unfolding: Model-Based Inspiration of Novel Deep Architectures (2014, methodology): “Sparse NMF (SNMF) [30] is used as a baseline, by optimizing the training objective, for each source, l .”
Archetypal analysis for machine learning (2010, methodology): “We found however, that the following projected gradient method inspired by the projected gradient method for NMF [20] using the normalization invariance approach of Eggert and Körner [12] worked efficiently in practice.”
Single-channel speech separation using sparse non-negative matrix factorization (2006, methodology): “We will use the sparse non-negative matrix factorization method proposed by Eggert and Körner [12] for both tasks.”
RolX: structural role extraction & mining in large graphs (2012, background): “It is worth pointing out that RolX can naturally incorporate other variants of matrix factorization such as imposing sparseness constraint on F and/or G by incorporating some regularization terms in the objective function [10]).”