fig4

Figure 4. Replacing MLPs in the output block of the MACE model^[50] with KANs employing B-spline basis functions. (A) The general framework of MACE model. z_i stands for the chemical species of atom i. $$ \vec{r}_{i} $$ indicates the atomic positions of atom i. $$ \vec{h}_{i} $$ represents the learnable features of atom i. (B) Replacing the linear combinations and MLPs in the output block with KANs. MLPs: Multi-layer perceptrons; KANs: Kolmogorov-Arnold Networks; MACE: Higher order equivariant message passing neural network.