In the last year, I have rediscover an interesting in machine learning, many years after the "Artificial Intelligence" course at the University of Pisa. So I took the chance the implement a few basic ML algorithms, like Logistic Regression and Factorization Machines, using this project as a playground to learn Python as well.
I believe this repo will be interesting for those who like to understand not only what works, but also why it works. Using a more established ML Library for a production workload is definitely a better idea (weka, mllib, sklearn and so on), but understand what's going on inside it is even better, in my opinion.
- Logistic Regression
- with Adagrad optimiser
- with Adadelta optimiser
- with FTRL Proximal optimiser
- Factorization Machine (Binary Classification)
- with Stochastic Gradient Descent optimiser
- with Adagrad optimiser
Code in this project is released under the Apache 2.0 license.
- http://scikit-learn.org/stable/developers/contributing.html#rolling-your-own-estimator
- https://github.com/kastnerkyle/kaggle-criteo/blob/master/clf.py
- https://gist.github.com/ceshine/f7f93046c58fe6ee840b
- http://www.eecs.tufts.edu/~dsculley/papers/ad-click-prediction.pdf
- https://github.com/fmfn/FTRLp/blob/master/FTRLp.py
- https://github.com/saiias/Adadelta/
- http://www.libfm.org
- https://github.com/srendle/libfm
- https://github.com/chenhuang-learn/ftrl
- https://www.kaggle.com/qqgeogor/ftrl-minibatch-campatible-with-sklearn