|
|
|
Figure 1
Schematic of the differentiable computational graph in MapsTorch. Inputs include elemental amplitudes and fitting parameters. The differentiable models include fluorescence peak lines, elastic and Compton scattering, and escape peaks. The sum (along with background) forms the model spectrum (Output), which is compared with the experimental spectrum via a loss/fitting quality function. AD supplies gradients to an optimizer such as Adam, closing the optimization loop. Blue arrows mean the forward process, red arrows mean the AD gradient calculation process, and the green arrow means the optimization process. |
Open
access
access
journal menu![[Figure 1]](gy5077fig1.jpg)



