Figure 2.

A schematic diagram of generating two kinds of eigenvectors. (a) the upper part: the similarity based feature vector: a n-dimensional disease-miRNA pair feature vector is obtained by the auto-encoder processing the dimension reduction of the cascaded m-dimensional disease feature vector and h-dimensional miRNA feature. (b) the lower part: the feature vector generated based on DeepWalk: a 2n-dimensional disease-miRNA pair feature vector is obtained by the cascaded n-dimensional disease feature vector and n-dimensional feature.