A semi-supervised learning approach to polymetallic nodule parameter modeling

摘要: We explore spatial modeling of polymetallic nodule parameters in the Clarion-Clipperton zone (CCZ) using a semisupervised learning approach. Spatial models that utilize factors affecting formation or their proxy variables, are useful tool for characterizing CCZ deposits and economic potential. Some such based on neural networks have been explored literature, but these approaches employed supervised learning. These rely hand-modeled features to incorporate effect topography, one key formation. employ unsupervised via auto-encoders principal component analysis efficiently capture local topography around point being modeled, express it terms few features. then use subsequent learningbased parameter modeling. Thus, this is overall show efficient incorporation bathymetric information into yields better performance as compared topographic