A Compressive Sensing Model for Speeding Up Text Classification

Algorithm 1

Flow of SRM sensing algorithm.

	Task : Perform Φ·X in which Φ is one of SRMs
	Input: The BOW feature matrix X = [x1,…,xi,…,xL1], the measurement number M, and a fast transform operator F(·).
	Main iteration: Iterate on i until i > L1 is satisfied.
(1)	Pre-randomization: randomize xi by uniformly permuting its sample locations. This step corresponds to multiplying xi with E.
(2)	Transform: apply a fast transform F(·) to the randomized vector, e.g, FFT, DCT, etc.
(3)	Subsampling: randomly pick up M samples out of N transform coefficients. This step corresponds to multiplying the transform coefficients with D.
	Output: The CS feature matrix Y = [y1,…,yi,…,yL1].