A REGULARIZED TENSOR NETWORK FOR CYCLONE WIND SPEED ESTIMATION

Maximum wind speed (MWS) is an important characteristic of tropical cyclone (TC). Estimation of MWS with remote sensing images of TCs via machine learning is a relatively new and challenging task. Here we propose a novel and effective method, Regularized Tensor Network (RTN), to estimate MWS using multispectral images (MSIs). RTN is a transductive regression model, built on a deep Tensor Network (TN) combined with two regularizations: manifold learning and categorization error. Experimental results showed that RTN outperformed several classic regression methods as well as advanced models based on deep learning.