EURASIP Journal on Audio, Speech, and Music Processing

Table 8 Some main differences and similarities of the PCA, NMF and SC methods

From: High level feature extraction for the self-taught learning algorithm

	PCA	NMF	SC
Number of bases	Less or same	Less, same or more	Less, same or more
wrt input dimension
Bases orthogonality	Yes	No	No
Bases learning	Analytic	Iterative	Iterative
Data representation	Linear	Non-linear	Non-linear
Data sign	Any	Positive	Any
Sparsity	No	Uncontrollable	Adjustable

Back to article page