This project implements the EQL÷ network for neural symbolic regression. The basis for this implementation are the paper by Martius and Lampert introducing the EQL network, and the subsequent work by Sahoo et al. introducing the EQL÷ network.
The script EQL_div.py implements the custom architecture of the EQL÷ network. formula_writer.py provides a function to write a learned network down as a symbolic formula. In model_selection.py, the method for selecting the optimal model among possible candidates is implemented. train_on_function.py provides a simple interface for tests, where a target function is given, which is learned using a given network structure.