diogenes

diogenes copied to clipboard

ConvergenceWarning - hitting max iterations and no convergence on 3gb data

8

python2.7/site-packages/sklearn/svm/base.py:209: ConvergenceWarning: Solver terminated early (max_iter=1000). Consider pre-processing your data with StandardScaler or MinMaxScaler. % self.max_iter, ConvergenceWarning)

johnsanterre

For example, say we have a number of different inspections on a number of different buildings, but there is more than one inspection per building. We may wish to aggregate...

zar1

array_emitter.py, line 965: The conditions should all be

kenben

Relook at proformance of utils.join

1

Re: James utils.join seems to perform poorly compared to similar functionality in R. The code looks like it might benefit from a re-write.

johnsanterre

i.e. | train features interval | train label interval | test features interval | test label interval | | --- | --- | --- | --- | | 2006-2007 |...

zar1

zar1

Consider diogenes in relation to Drain

1

https://github.com/dssg/drain It might be worth pulling in code, writing some docstrings, and incorporating into diogenes.

zar1

Include templates for things that our users might want to do.

zar1

reporting -

1

plot accuracy on subsets, ie, avg. acc. on subsets of size [5,10,15, 25, 50] etc.

johnsanterre

Like this one: https://github.com/dssg/police-eis/tree/master/evaluation/webapp. Also, export experiment to the directory of pkl files that the webapp expects.

zar1