Low-error two-source extractors for polynomial min-entropy

The aim of this paper was to present a new two-source extractor for min-entropy nα for some α < 1 having exponentially-small error. Say we are given two independent sources X and Y . The technique was to condense Y using a somewhere-random condenser to a somewhere-random source with t rows, where one of the rows has a high min-entropy rate. Then, in order to break correlations between the rows, we would use an advice correlation breaker (following the work of [2] and also techniques developed in [3], for example). The advice correlation breaker that we constructed may be of independent interest in case the number of rows is small. Specifically: