Ionosphere Tomographic Model Based on Neural Network with Balance Cost and Dynamic Correction Using Multi-Constraints

A Neural network (NN) is a promising tool for the tomographic inversion of ionosphere. However, existing research has adopted an unbalanced cost function training purposes and preset image constraint purposes, resulting in output being dominated by measurements. To address these problems, we proposed NN-based model with balance dynamic correction process (BCDC) ionosphere inversion. The composed two terms corresponding to measurements selected constraints, respectively. produced forward NN corrected dynamically fitting each vertical profile orthogonal basis functions (EOFs) Chapman then smoothing voxels layer moving window approach horizontally. used calculate slant total electron content (STEC) parameter, which further translated into term horizontal constraints. Experiments were carried out validate BCDC method compared recently developed international reference (IRI) model. Results show that parameters derived from demonstrate good consistency observations. Comparing methods, performs better validations profiles, F2 peak density (NmF2), STEC parameter map. Further analysis also shows benefit achieve quality.