Links to original paper published by IEEE Computer Society : [1], [2]
Link to Netflix Dataset Used : [1]
-
Presentation.pdf : Explains the paper. Was written in Latex Beamer, tex code is in presentation.tex
-
recommender_final.py : The final recommender. Includes biases and regularization. Requires mf.py to be imported to run. Use directly on any dataset by changing line 19 in recommender_final.py.
-
recommender_final_toy_dataset.py shows how exactly Matrix Factorization Techniques work by considering a 5x5 toy dataset.
-
The .ipynb_ files include visualizations of RMSE decreasing with iterations when fitting on the training dataset. All .ipynb files are standalone and do not require importing mf.py
-
feasible_data_n.txt : Files with only the first n datapoints from whole dataset. Used for Testing.
-
Training and Testing Data :
Not given separately. Program randomly separates k% of data as Test data, trains on remaining, then tests on the k% values. Default k=20, can be changed on line 154.
