kimianoorbakhsh / symbolic-math

(Symbolic Math) Reimplementation of Deep Learning for Symbolic Mathematics

A reimplementation of Lample & Charton (2019) Deep Learning for Symbolic Mathematics

• arxiv
• pdf
• Open Review

Codebase guide

1. generate random math expressions in binary tree form (random_trees.py)
2. map tree to prefix (random_trees.py)
3. prefix to infix (infix_prefix.py)
4. infix to prefix (infix_prefix.py)

Workflow

1. backward_generation.ipynb - generate trees, generate target using sympy, simplify, make sequence (input & target)

Can also run

$ python generate_dataset.py --cpu 12 --num 10000 --n 8

2. seq2seq_model.ipynb basic model

3. transformer.ipynb trains model in google cloud.