This code can be used to reproduce the numerical experiments in the paper "Generalizing Outside the Training Set: When Can Neural Networks Learn Identity Effects?" by S. Brugiapaglia, M. Liu, and P. Tupper (Proceedings of CogSci 2020).
Full paper here: https://arxiv.org/abs/2005.04330
Clone this repo
$ git clone https://github.com/mattjliu/Identity-Effects-Cogsci-2020.git
Install dependencies
$ pip install -r requirements.txt
Results and training history can be found in the runs
folder. Experiments and results can be viewed in experiments.ipynb
.