In this project, the MovieLens dataset was used to classify movies and recommend the best option per user. The repo is organised as follows:
dataset_preparation.ipynb
: Jupyter notebook that does data preprocessingmatrix_factorization.ipynb
: Jupyter notebook that uses the embeddings technique provided by Pytorch framework to factorize the tabular matrix created by the data preprocessing notebookmlp.ipynb
: Jupyter notebook where the multi - layer perceptron is implementedu.data
: The dataset used in the project
In this project, there was an experiment where matrix factorization was both carried out by a multi - layer perceptron and an embeddings neural network. The result was much better using the embeddings neural network. Finally, there is a Jupyter cell where a custom keras - like neural network fitting was implemented. This cell displays live progress of the neural network while it's training.