forougha / neuralmath

Combining Symbolic and Function Evaluation Expressions In Neural Programs

Combining Symbolic and Function Evaluation Expressions In Neural Programs

This repository contains the code for this paper

File equationGenerator.ipynb is an iPython notebook that describes the dataset generation method proposed in the paper, and lstmTreeNotebook.ipynb contains the code for training over the generated equations.

For a PyTorch implementation of the models in the paper and our recently released memory-augmented tree-LSTM architecture checkout here

neuralmath's Issues

Python3 Compatability

I cannot get this repo to run under a python3 system and I failed at setting up a python2 as sympy does not support python2.7 anymore.

do you happen to have an updated version of this repo, I'd like to generate some equations of my own?

what is black-box function evaluations?

In your paper, you mention "black-box" 13 times and "black-box function evaluations" 8 times without providing description about them. And it is also listed as one contribution of your paper, could you help to elaborate about this?