A dimensionality reduction technique using random matrices to efficiently approximate high-dimensional data with linear complexity.