Histogram of the number of learning sample galaxies in the cells as encountered by the 50 million queries to the LUT. The cells of the LUT that are most queried by the GREAT10 data are sufficiently filled by the learning sample. Using a learning sample of 9 million galaxies, we find that only 5% of the queries found less than 5 learning galaxies in their cells. Note that for those queries falling on an empty cell, our implementation returns the mean correction from the cells neighbors.