Weighted t-SNE

Welcome!

How to use

To configure all the hyperparameters of Weighted t-SNE, you only need to create a config.py file. An example can be downloaded here. It also contains the necessary documentation. To set the weights of each features you should use a .csv file as in this example.

You will need Python 3 to run this code. Check if the needed libraries are installed with:

python3 check_dep.py

And for the weighted t-SNE visualization, run:

python3 wtsne.py config.py

Datasets

You can download the datasets used in the experiments here and the scores from the feature scorers here.

The complete CuMiDa can be found here: https://sbcb.inf.ufrgs.br/cumida

Experiments

You can download the complete configuration of the experiments here.

Interactive plots of the results can be seen here: https://sbcblab.github.io/wtsne

Libraries

This implementation of relevance aggregation uses the following Python 3.7 libraries:

• NumPy
• Pandas
• Matplotlib
• SciPy
• Scikit-learn
• Plotly
• openTSNE
• Relevance Aggregation

How to cite

If you use our code, methods, or results in your research, please consider citing the main publication of weithed t-SNE:

To be published.

Bibtex entry:

@article{grisci2021relevance,
  title={},
  author={},
  journal={},
  year={},
  doi = {},
  publisher={}
}

Contact information

• Bruno I. Grisci - PhD candidate (Institute of Informatics - UFRGS)

  • [email protected]

• Prof. Dr. Mario Inostroza-Ponta - Associate Professor (Departamento de Ingeniería Informática - USACH)

  • [email protected]

• Prof. Dr. Marcio Dorn - Associate Professor (Institute of Informatics - UFRGS)

  • [email protected]

• http://sbcb.inf.ufrgs.br/