amndeep7 / heterogeneous_groups Goto Github PK

add something that reads in a config file that specifies the args to the grouper class and make it be able to read the data from a file

dockerize it as well

use cmdline args to specify file names

make all that code health + test stuff as github actions pipelines

see about how to do it as pre-commit stuff as well

make sure to have it pass black, pylint, safety, bandit, isort, mypy

maybe stuff can be separated out? not sure

make it easier to use by figuring out the init.py stuff

add actual how to use section

add actual development/contribution instructions

this is getting pretty obnoxiously long - have it be able to take in as args the datasets and algorithm types

just wanna compare the different algos to see how each shapes up and verify my asssumptions about runtime

Code is fine to be wider, but the built in documentation displayer thing that coc.nvim does only shows up to 80c I think, maybe even 79

currently bound to 3.9, I think it might be possible to go down to 3.8 or even 3.7

figure out how to get something up on pypi

make github autorelease to there as well as make a package for here + releases/tags

make every possible combination of (num_items)C(num_items/num_g) - if non-zero modulo then also (num_items/num_g)+1

out of that set of combinations, make every possible combination of size num_g that uniquely includes every item

applying the same formula to get the collective dissimilarity per group (heterogeneity measure?), find the combination that maximizes that value across all groups

this can be its own algorithm but i'm presuming that it will be considerably slower than my original algorithm which in and of itself will probably be considerably slower than the paper's algo

be better

move that stuff outta the actual file and into the tests directory

add more

currently (c) me but ideally wanna put it under a useful license and also need to figure out what to do w/r to the algorithm which my implementation is a derivative work of (like it follows the idea but there are numerous differences such as the null stuff being removed , categories being added, etc.)

using pydantic but gotta add that func to verify stuff like how upper > lower in boundaries, etc.