Catalytic Role Of Noise And Necessity Of Inductive Biases In The Emergence Of Compositional Communication

Communication is compositional if complex signals can be represented as a combination of simpler subparts. In this paper, we theoretically show that inductive biases on both the training framework and the data are needed to develop a compositional communication. Moreover, we prove that compositionality spontaneously arises in the signaling games, where agents communicate over a noisy channel. We experimentally confirm that a range of noise levels, which depends on the model and the data, indeed promotes compositionality. Finally, we provide a comprehensive study of this dependence and report results in terms of recently studied compositionality metrics: topographical similarity, conflict count, and context independence.

[1]  M. Dufwenberg.  Game theory. , 2011, Wiley interdisciplinary reviews. Cognitive science.

[2]  Zenon W. Pylyshyn,et al.  Connectionism and cognitive architecture: A critical analysis , 1988, Cognition.

[3]  Brian Skyrms,et al.  Self-assembling Games , 2017, The British Journal for the Philosophy of Science.

[4]  Alexander Peysakhovich,et al.  Multi-Agent Cooperation and the Emergence of (Natural) Language , 2016, ICLR.

[5]  Stephen Clark,et al.  Emergence of Linguistic Communication from Referential Games with Symbolic and Pixel Input , 2018, ICLR.

[6]  Simon Kirby,et al.  Spontaneous evolution of linguistic structure-an iterated learning model of the emergence of regularity and irregularity , 2001, IEEE Trans. Evol. Comput..

[7]  Michael Oliphant,et al.  Learning and the Emergence of Coordinated Communication , 1997 .

[8]  Michael Cogswell,et al.  Emergence of Compositional Language with Deep Generational Transmission , 2019, ArXiv.

[9]  José M. F. Moura,et al.  Natural Language Does Not Emerge ‘Naturally’ in Multi-Agent Dialog , 2017, EMNLP.

[10]  Bernhard Schölkopf,et al.  Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations , 2018, ICML.

[11]  Simon Kirby,et al.  Understanding Linguistic Evolution by Visualizing the Emergence of Topographic Mappings , 2006, Artificial Life.

[12]  Piotr Milos,et al.  Emergence of compositional language in communication through noisy channel , 2020 .

[13]  Yi Ren,et al.  Inductive Bias and Language Expressivity in Emergent Communication , 2020, ArXiv.

[14]  Jason Lee,et al.  Emergent Translation in Multi-Agent Communication , 2017, ICLR.

[15]  Eugene Kharitonov,et al.  Compositionality and Generalization In Emergent Languages , 2020, ACL.

[16]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[17]  Julian Zubek,et al.  Developmentally motivated emergence of compositional communication via template transfer , 2019, ArXiv.

[18]  Andriy Mnih,et al.  Disentangling by Factorising , 2018, ICML.

[19]  Abhinav Gupta,et al.  Towards Graph Representation Learning in Emergent Communication , 2020, ArXiv.

[20]  R. Kirk.  CONVENTION: A PHILOSOPHICAL STUDY , 1970 .

[21]  Sang Joon Kim,et al.  A Mathematical Theory of Communication , 2006 .

[22]  W. Strange.  Evolution of language. , 1984, JAMA.

[23]  Ben Poole,et al.  Categorical Reparameterization with Gumbel-Softmax , 2016, ICLR.

[24]  Vladimir I. Levenshtein,et al.  Binary codes capable of correcting deletions, insertions, and reversals , 1965 .

[25]  Nando de Freitas,et al.  Compositional Obverter Communication Learning From Raw Visual Input , 2018, ICLR.

[26]  Martin Wattenberg,et al.  How to Use t-SNE Effectively , 2016 .

[27]  Andrew M. Dai,et al.  Capacity, Bandwidth, and Compositionality in Emergent Language Learning , 2020, AAMAS.

[28]  Stefan Lee,et al.  Learning Cooperative Visual Dialog Agents with Deep Reinforcement Learning , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[29]  Pieter Abbeel,et al.  Emergence of Grounded Compositional Language in Multi-Agent Populations , 2017, AAAI.

[30]  Nando de Freitas,et al.  Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning , 2018, ICML.

[31]  Nikolaus Kriegeskorte,et al.  Frontiers in Systems Neuroscience Systems Neuroscience , 2022 .

[32]  Julian Zubek,et al.  Measuring non-trivial compositionality in emergent communication , 2020, ArXiv.

[33]  Jonathan Berant,et al.  Emergence of Communication in an Interactive World with Consistent Speakers , 2018, ArXiv.

[34]  Matthew E. Taylor,et al.  A Very Condensed Survey and Critique of Multiagent Deep Reinforcement Learning , 2020, AAMAS.

[35]  Christopher Burgess,et al.  beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework , 2016, ICLR 2016.

[36]  Joshua B. Tenenbaum,et al.  Building machines that learn and think like people , 2016, Behavioral and Brain Sciences.

[37]  Kaiming He,et al.  Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour , 2017, ArXiv.

[38]  Shimon Whiteson,et al.  Learning to Communicate with Deep Multi-Agent Reinforcement Learning , 2016, NIPS.

[39]  Samy Bengio,et al.  Discrete Autoencoders for Sequence Models , 2018, ArXiv.

[40]  Geoffrey E. Hinton,et al.  Visualizing Data using t-SNE , 2008 .

[41]  Gintare Karolina Dziugaite,et al.  Linear Mode Connectivity and the Lottery Ticket Hypothesis , 2019, ICML.

[42]  Geoffrey E. Hinton,et al.  Semantic hashing , 2009, Int. J. Approx. Reason..

[43]  Michael Bowling,et al.  Ease-of-Teaching and Language Structure from Emergent Communication , 2019, NeurIPS.

[44]  Marco Baroni,et al.  How agents see things: On visual representations in an emergent language game , 2018, EMNLP.

[45]  Simon Kirby,et al.  Compositional Languages Emerge in a Neural Iterated Learning Model , 2020, ICLR.

[46]  Abhinav Gupta,et al.  Exploring Structural Inductive Biases in Emergent Communication , 2020, ArXiv.

[47]  T. Deacon,et al.  Language Development From an Ecological Perspective: Ecologically Valid Ways to Abstract Symbols , 2018 .

[48]  Chelsea Finn,et al.  Language as an Abstraction for Hierarchical Deep Reinforcement Learning , 2019, NeurIPS.