Table 2.
Summary of the advantages of SVD and NMF matrix factorization methods. The advantages of one method complement the disadvantages of the other (Langville et al. 2006). For example, the NMF optimization problem is nonconvex with local minima resulting in solutions that depend on the initialization of the algorithm.
SVD advantages | NMF advantages |
---|---|
Optimal rank r approximation | Results are nonnegative |
Fast to compute | Results are sparse |
Unique | Sparsity and nonnegativity lead to improved interpretability |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.