Workflow

The general SALTED workflow is summarised below. Detailed descriptions of each step will given in the following sections of the tutorial.

SALTED workflow

  1. Calculate electron density and density fitting (DF) coefficients.
  2. Generate (optionally sparse) \(\lambda\)-SOAP descriptors by rascaline, and sparsify the atomic environments by farthest point sampling (FPS) method. This step uses the salted functions initialize, sparse_selection, and sparse_descriptor.
  3. Calculate RKHS related quantities, including kernel matrix \(\mathbf{K}_{MM}\), associated projectors, and the feature vector \(\mathbf{\Psi}_{ND}\). This step uses the salted functions rkhs_projector and rkhs_vector.
  4. Optimize GPR weights by either direct inversion or CG method, and save the optimized weights. This uses either the functions hessian_matrix and solve_regression, or minimize_loss.
  5. Validate the model using the validation function.
  6. Predict density fitting coefficients of new structures using the GPR weights obtained in the previous step, saving the predicted density coefficients. This uses the prediction function.
  7. Use the predicted density coefficients to calculate properties derived from the predicted electron density.