Semi-Supervised Cross-Modal Retrieval Based on Discriminative Comapping

Algorithm 1

Optimization for MRRDC.

	Input: all image feature matrices , all text feature matrices , and the corresponding semantic matrix .
	Initial: , and set the parameters and maximum iteration time. is the step size in the alternating updating process, and is the convergence condition.
	Repeat:
	Until
	Repeat:
	Until
	Until maximum iteration number
	Output: