Fast Emulation of Self-organizing Maps for Large Datasets