Curve approximation based on Neural Networks for general classes of totally positive rational bases

In this repository, the code for the experimentation is provided. Different curves were approximated using a one-hidden-layer neural network that uses rational basis as activation functions. Different rational basis were used.

1. In Experiments_table the code for Table 1 is provided.
2. The architecture and training implementation is described in the jupyter notebook. (Can be seen in markdown version here).

The implementation uses Tensorflow 2.0.