Hamish
Hamish
Hmmm never hit that before. I suspect it's the matrix calc. Have you verified the data is ok on smaller sets of the same data?
Yeah it's possibly just a limitation of this approach. As that matrix gets larger the computation will blow out. The matrix calcs are outside of this repo, so unfortunately it...
@olimpias yes of course!
@olimpias yeah i guess in general each variable needs to be plotted against Y. Potentially it would also be useful to see X1 against Xn, but that's less important.
Oh nice. Yeah you can train with a dictionary, there is an example in the tests. This pkg is used in a git tool somewhere also, I can't remember how...
interesting, have not looked at this for a long time. Will ping some others. Might be worth adding a test in a pull request, even if it is failing, will...
a threshold greater than zero requires more training to determine the validity of text.
@abraztsov hadn't planned on it, but we do accept PRs if you're interested in adding it?
happy to accept a PR for that if you want to add it? It should be a pretty small change.
FYI has been added in https://github.com/sajari/docconv/pull/54