The overall objective of this project is to improve the current CF module in mlpack to provide better rating prediciton, fast execution, and flexible APIs.
To be more specific, the objectives include:
- Solve the current problems which are affecting the accuracy (e.g. lack of data normalization, lack of alternative methods for aggregation of neighbor ratings).
- Add CF models which are more expressive, and models which can take implicit feedbacks into consideration (e.g. BiasSVD, SVD++).
- Benchmark the cf module with public datasets (e.g. MovieLens, Netflix).
- Accelerate program execution by identifying and modifying the codes which are slowing down the execution.