Paper: GLocal-K: Global and Local Kernels for Recommender Systems.
The proposed matrix completion framework based on global and local kernels, called GLocal-K, includes two stages: 1) pre-training an autoencoder using the local kernelised weight matrix, and 2) fine-tuning the pre-trained auto encoder with the rating matrix, produced by the global convolutional kernel. This repo provide the benmark with the processed data from Movie Recommender Systems.
Download this repository. As the code format is .ipynb, there are no settings but the Jupyter notebook with GPU.
- Use the data in dir
processed_data_for_matrix_completion
or create a csv file with the same format. - Insert the data path in the main code.
- Run the notebook and see the result.
- numpy
- scipy
- torch
More about the method can be found here
Download this repository. As the code format is .ipynb, there are no settings but the Jupyter notebook with GPU.
- Use the data in dir
processed_data_for_matrix_completion
or create a csv file with the same format. - Insert the data path in the main code.
- Run the notebook and see the result.
- numpy
- scipy
- torch
- Training: run script .heterogenrous-model/train.py
- Evaluating: run script .heterogenrous-model/eval.py
- torch
- scikit-learn
- transformers
Generate the sampled data in either csv or jsonlines format with the sample.ipynb.