A Deep Fusion Gaussian Mixture Model for Multiview Land Data Clustering

Figure 2

The computing paradigm of the multiview fusion Gaussian mixture model. Modality-specific encoders, modality-common encoder-decoder, and modality-specific decoders are linked in a cascaded manner where data are transferred into hidden representations of each view by modality-specific encoders; then, those hidden representations are concentrated, which are reconstructed via the modality-common encoder-decoder, and finally, the reconstructed hidden representations are decoded into the original data space by modality-specific decoders.