PatchGame: Learning to Signal Mid-Level Patches in Referential Games

Abstract

We study a referential game (a type of signaling game) where two agents communicate with each other via a discrete bottleneck to achieve a common goal. In our referential game, the goal of the speaker is to compose a message or a symbolic representation of "important" image patches, while the task for the listener is to match the speaker's message to a different view of the same image. We show that it is indeed possible for the two agents to develop a communication protocol without explicit or implicit supervision. We further investigate the developed protocol and show the applications in speeding up recent Vision Transformers by using only important patches, and as pre-training for downstream recognition tasks (e.g., classification).

PDF NeurIPS OpenReview Code Semantic Scholar

Cite

Text

Gupta et al. "PatchGame: Learning to Signal Mid-Level Patches in Referential Games." Neural Information Processing Systems, 2021.

Markdown

[Gupta et al. "PatchGame: Learning to Signal Mid-Level Patches in Referential Games." Neural Information Processing Systems, 2021.](https://mlanthology.org/neurips/2021/gupta2021neurips-patchgame/)

BibTeX

@inproceedings{gupta2021neurips-patchgame,
  title     = {{PatchGame: Learning to Signal Mid-Level Patches in Referential Games}},
  author    = {Gupta, Kamal and Somepalli, Gowthami and Anubhav, Anubhav and Hewa, Vinoj Yasanga Jayasundara Magalle and Zwicker, Matthias and Shrivastava, Abhinav},
  booktitle = {Neural Information Processing Systems},
  year      = {2021},
  url       = {https://mlanthology.org/neurips/2021/gupta2021neurips-patchgame/}
}