Simultaneous Dimensionality Reduction for Extracting Useful Representations of Large Empirical Multimodal Datasets