NCM-MDGNN: Data-driven graph neural networks guide codoping for stable Ni-rich layered cathodes

The design patterns and data flow changes of each link are shown in Figure 1. Specifically, the data preprocessing process involves the fusion of small sample data, the multi-ML ensemble algorithms, and the Kalman filter denoising algorithm to achieve precise sample enhancement and data preprocessing.

Secondly, a vectorization method is employed to construct a multi-stage subdivided spherical structure from the 0th level icosphere spherical surface. Meanwhile, we construct a three-dimensional spherical structure network with different level in a nested structure. The graph relationship matrix (including the vertex matrix and edge matrix) is recorded based on the relationship between vertices and edges.