We have obtained efficient and robust solutions for the problem of recall and training with missing data. Experimental results verified our method. All of our results can easily be generalized to the case of noisy inputs.
