############################################################################### ############################################################################### #### #### #### ACCURATE, AFFORDABLE, AND GENERALISABLE MACHINE LEARNING #### #### SIMULATIONS OF TRANSITION METAL X-RAY ABSORPTION SPECTRA #### #### USING THE XANESNET DEEP NEURAL NETWORK #### #### #### #### C. D. Rankine ~ T. J. Penfold #### #### #### ############################################################################### ############################################################################### ####################### # ARCHIVE DESCRIPTION # ####################### The dataset directory `datasets` contains nine first-row transition metal (Ti--Zn) dataset subdirectories; each contains three datasets: `default`, `preconv`, and `totalconv`. The first (`default`) contains the absorption cross-sections as calculated in FDMNES without any preprocessing. The second (`preconv`) contains the absorption cross-sections convoluted with a simple Lorentzian (details are in the SI for this Article); this is the dataset used throughout. The third (`totalconv`) contains the absorption cross-sections convoluted with an arctangent convolutional model (details are in the SI for this Article). The three dataset directories - `default`, `preconv`, and `totalconv` - contain four subdirectories each: `xyz_train`, `xanes_train`, `xyz_test`, and `xanes_test`. The xyz (X) directories contain the samples: the Cartesian coordinates in .xyz format. The XANES (Y) directories contain the labels: the XANES spectra in .txt format (matching the typical FDMNES output formatting). ####################### # ARCHIVE STRUCTURE # ####################### > datasets > ti > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > v > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > cr > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > mn > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > fe > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > co > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > ni > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > cu > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test > zn > default > xanes_train > xanes_test > xyz_train > xyz_test > preconv > xanes_train > xanes_test > xyz_train > xyz_test > totalconv > xanes_train > xanes_test > xyz_train > xyz_test