We need to add the Clebsch-Gordan (CG) utils somewhere, so we decided to add it to rascaline.
Implementation
There are severe problems with all existing CG utils which need to be addressed however. @Luthaf had some thoughts about it.
Mathematical documentation
As far as I know the most recent version of the recursive relationship of ACDC features is in the supplementary of the "Optimal radial basis for density-based atomic representations" paper.
Link: https://pubs.aip.org/jcp/article-supplement/1013253/zip/104106_1_supplements/
See Section VII. A SIMPLIFIED DERIVATION OF THE NICE ITERATION
@ecignoni is working on simplified math paper that explains the key points important to derive the equivariant representations that might be more useful for a doc page in rascaline, since the before-mentioned supplementary is quite technical. Will give an update in two weeks
We need to add the Clebsch-Gordan (CG) utils somewhere, so we decided to add it to rascaline.
Implementation
There are severe problems with all existing CG utils which need to be addressed however. @Luthaf had some thoughts about it.
Mathematical documentation
As far as I know the most recent version of the recursive relationship of ACDC features is in the supplementary of the "Optimal radial basis for density-based atomic representations" paper. Link: https://pubs.aip.org/jcp/article-supplement/1013253/zip/104106_1_supplements/ See Section VII. A SIMPLIFIED DERIVATION OF THE NICE ITERATION
@ecignoni is working on simplified math paper that explains the key points important to derive the equivariant representations that might be more useful for a doc page in rascaline, since the before-mentioned supplementary is quite technical. Will give an update in two weeks