/simple_logistic

Playing with building things just using numpy.

Primary LanguageJupyter Notebook

A super basic Logistic regressor built just using numpy

logistic/logistic.py

this file contains the entire implementation for the logistic regression along with the SGD based optimization. The LogisticRegression class encapsulates everything.

main_driver.py

If you do not want to go through a notebook and just test the logistic regression, you can just run the this file and it will do a 5 fold cross validation with a L2 regularization. You should get an AUC of around 0.75 and an F1-score of around 0.75

dependencies.txt

List of dependencies if you want to run it in a virtual env. But I think all you need are pandas , numpy, and sklearn. All the others are pulled in due to jupyter.