Resources

Papers, articles, and other material

Modeling

• “Emergence of Grounded Compositional Language in Multi-Agent Populations” by Igor Mordatch, Pieter Abbeel. Mordatch and Abbeel demonstrate the emergence of a language with relevant, materially grounded noun, adjective, and verb symbols from a simple reinforcement learning multi-agent collaboration task.
  • See slides from a JC presentation given on this paper.
• “On the Spontaneous Emergence of Discrete and Compositional Signals” by Nur Geffen Lan, Emmanuel Chemla, Shane Steinert-Threlkeld. Lan et al. demonstrate that a signalling language games can be understood as an autoencoder problem and demonstrate that discrete messages naturally arise within these messages from training autoencoders in such contexts.
• “Learning Emergent Discrete Message Communication for Cooperative Reinforcement Learning” by Sheng Li, Yutai Zhou, Ross Allen, Mykel J. Kochenderfer.
• “Compositional Observer Communication Learning from Raw Visual Input” by Edward Choi, Angeliki Lazaridou, Nando de Freitas.
• “An Improvement for Capsule Networks using Depthwise Separable Convolution” by Nguyen Huu Phong, Bernardete Ribeiro.
  • Looking into using this as our visual unit for Siamese architecture.

Computational Linguistics

• “Determining Compositionality of Word Expressions Using Various Word Space Models and Measures”

More research and resources forthcoming.