Linear Discriminative Learning for Image Classification

Table 4

Running time comparison on the MNIST dataset for various feature dimensions using eigen and random projections.

	Dimension
	Eigen-20		Eigen-54		Eigen-120		Random-20		Random-54		Random-120
Method	Recog.	Time	Recog.	Time	Recog.	Time	Recog.	Time	Recog.	Time	Recog.	Time
CRC	71.61%	0.73	73.00%	0.84	73.89%	1.05	53.49%	0.54	66.61%	0.61	71%	0.80
SRC	76.71%		77.28%		74.85%		59.42%		70.71%		72.85%
ALWCR	77.71%		78.28%		76.85%		63.14%		74.28%		75.28%

Time = running time in seconds for all testing samples in the dataset. The bold numbers are the highest recognition rates and lowest running time for experiment.