On the interaction between supervision and self-play in emergent communication

Authors: Ryan Lowe*, Abhinav Gupta*, Jakob Foerster, Douwe Kiela, Joelle Pineau

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We then empirically investigate various S2P schedules that begin with supervised learning in two environments: a Lewis signaling game with symbolic inputs, and an image-based referential game with natural language descriptions.
Researcher Affiliation Collaboration Ryan Lowe , Abhinav Gupta MILA Jakob Foerster, Douwe Kiela Facebook AI Research Joelle Pineau Facebook AI Research MILA
Pseudocode No The paper describes methods and processes but does not include any explicitly labeled 'Pseudocode', 'Algorithm', or structured algorithmic blocks.
Open Source Code Yes 1Code is available at https://github.com/backpropper/s2p.
Open Datasets Yes For this game, the target language L is English we obtain English image descriptions using caption data from MS COCO and Flickr30k.
Dataset Splits Yes To help describe these methods, we further split the seed dataset D into Dtrain, which is used for training, and Dval which is used for early-stopping.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory used for running the experiments.
Software Dependencies No The paper mentions software components and models like 'GRU', 'Gumbel-Softmax', 'Resnet50', and optimizers 'Adam, SGD', but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes We provide hyperparameter details in Table 1.