The robustness to noise and outliers is an important issue in linear representation in real applications. We focus on the problem that samples are grossly corrupted, which is also the " sample specific " corruptions problem. A reasonable assumption is that corrupted samples cannot be represented by the dictionary while clean samples can be well represented. This assumption is enforced in this p...

Abstract We propose a novel statistical inference methodology for multiway count data that is corrupted by false zeros are indistinguishable from true zero counts. Our approach consists of zero-truncating the Poisson distribution to neglect all values. This simple truncated dispenses with need distinguish between and counts reduces amount be processed. Inference accomplished via tensor completi...

For TREC retrieval from corrupted data was studied through retrieval of single target documents from a corpus which was corrupted by producing page images corrupting the bit maps and applying OCR techniques to the results In general methods which attempted a probabilistic estimation of the original clean text fare better than methods which simply accept corrupted versions of the query text