Pytorch implementation of the paper 'Compositional language emerge in a neural iterated learning' (ICLR 2020).
Hey:
Thanks for your interest in our work. We apologize that this is a very fundamental version of our work. But we promise to update it latter. However, due to the time constraints (the authors are fascinated by another topic related to iterated learning and do not want to spend too much time on this task :-( ), the update might come later, as we plan to combine different applications of iterated learning and release a thorough code or toolkit for it.
We guarantee that this initial version code can work when you installed the required packages (like numpy, torch 1.0, etc...) and the essential part of our algorithm could be verified. You can run the "train.py" directly, and see the topological similarity indeed increased to nearly 1.0 after some generations.
Finally, if you have any concerns about our work, please do not hesitate to send us an email to renyi.joshua@gmail. We're ready and glad to answer all your questions.
Thanks very much.
Authors