A statistical technique that simultaneously partitions variability of both X and Y Variables (or matrices), somewhat similar to Principal Components. A PLS model attempts to identify the multidimensional direction in the X space that explains maximum multidimensional Variance direction in the Y space.