Local environment-induced atomic features (LEAFs).
This data has been generated and used as described in the manuscript:

A. Vasylenko, S. Schewe, L. Daniels, J. Claridge, M. Dyer, M. J. Rosseinsky,
"Learning atoms from crystal structure"

1. LEAFs.csv: 
A collection (a dataframe) of LEAFs for chemical elements,
which could be used directly for digitalization of chemical elements and their representation as vector,
and for modelling materials as compositions.

The schematics of computing LEAFs from crystal structures reported in Inorganic Crystal Structure Database (ICSD)
is presented in Figure 1, of the manuscript.

Figure 2, 3 and 4 illustrate the usage of data in LEAFs.csv in different tasks.

2. leaf+.csv:
An extended collection of descriptors for chemical elements, which is obtained by concatenating LEAFs with Magpie
(H. Ucar, D. Paudyal, K. Choudhary, Comput. Mater. Sci., 209 (2022), 111414) 

Employing leaf+, e.g., produces accurate results for crystal prediction discussed in the manuscript, and presented
in Figure 2 for LEAFs (See structure_predicted.csv) 

3. LEAFonehot_decimals3_nozero.pickle:
The full Matrix of local structure environment, illustrated schematically in Figure 5.
This data can be used a source for machine learning of elemental representation, including in the end-to-end scheme
in connection with the materials property modelling.