Cimagroup / Experiments-Representative-datasets

Representative Datasets

Gonzalez-Diaz, R., Gutiérrez-Naranjo, M.A. & Paluzo-Hidalgo, E. Topology-based representative datasets to reduce neural network training resources. Neural Comput & Applic 34, 14397–14413 (2022). (Paper)

Three experiments were developed:

1. Iris dataset experiment,
2. Digits dataset experiment.
3. Different synthetic datasets.

In all of them three sets were considered, the original dataset, the dominating dataset and a random dataset. Besides, the Algorithm based in proximity graphs and dominating sets was implemented and can be found in the auxiliary_fun.py.

List of main needed libraries

• ripser
• keras
• gudhi
• Giotto-tda

Experiments were run on a computer with Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz

An old preprint version of the paper can be found (here).