Linear/Non-Linear methodologies for reduced dimensionality

With the exponential advancement in the linear and non-linear methodologies, there has been no specific comparative study across methods, which is known to be conducted, on common dataset with common dimensionality factors. This study will help understand the impact of linear/nonlinear methods, across different dimensions of latent space, on the quality of features generation in collaborative filtering. As a result, this thesis aims to explore, generalise and consolidate the impact of different methodologies across common variation of dimensions on publicly available datasets. This experimentation forms the foundation of the thesis. The implementation of code is present in the mentioned below .pynb files.