Input: Series with COO format, so the data as "flags" with multiindex of row & column IDs (does this need to be Series or is Dataframe ok?). This represents a Levi graph where there are "nodes" of 2 categories of information (ex. - tokens and documents). Eventually, grabble will support handling data with more than two categories, and any two can be selected for the Levi graph (ex. - tokens and date, document category and tokens, etc)
Goal: Incidence structure (ex - something like a doc-term matrix) that can be used for more calculations, including compatibility with networkx, does not use pandas's built-in sparse accessor
Tasks:
[ ] Accessor requires custom validation function. Probably use beartype
[ ] Property: incidence structure
[ ] Functions: ?? networkx stuff, linear algebra stuff
@tbsexton Just want to check that this is the correct "input" format for the accessor (multi-index pandas Series or dataframe)? Then from here we would use df.levi.foo to get out the data in various matrix formats?
Create Levi Pandas accessor.
Input: Series with COO format, so the data as "flags" with multiindex of row & column IDs (does this need to be Series or is Dataframe ok?). This represents a Levi graph where there are "nodes" of 2 categories of information (ex. - tokens and documents). Eventually,
grabble
will support handling data with more than two categories, and any two can be selected for the Levi graph (ex. - tokens and date, document category and tokens, etc)Goal: Incidence structure (ex - something like a doc-term matrix) that can be used for more calculations, including compatibility with
networkx
, does not use pandas's built-in sparse accessorTasks:
beartype