Skip to main content
Version: Next

MatPL Tutorials

👉Source repository

MatPL includes 8 descriptor types with translational, rotational, and permutational invariance:

        1. 2-body(2b)
        2. 3-body(3b)
        3. 2-body Gaussian(2b gauss)
        4. 3-body Cosine(3b cos)
        5. Moment Tensor Potential(MTP)
        6. Spectral Neighbor Analysis Potential(SNAP)
        7. DP-Chebyshev(dp1)
        8. DP-Gaussian(dp2)

4 trainable model types:

        1. Linear
        2. Neural Network(NN)
        3. DP se_e2_a(Pytorch)
        4. Neuroevolution Potential (NEP) (supports multi-GPU, multi-node training and ultra-fast large-scale LAMMPS MD)

2 efficient training optimizers:

        1. Adaptive Moment Estimation (ADAM)
        2. Reorganized Layer Extended Kalman Filtering (LKF)