2026-04-17T16:56:05Zhttps://recercat.cat/oai/request

oai:recercat.cat:10230/526232025-12-24T08:34:04Zcom_2072_6col_2072_452952

Interpretable agent communication from scratch (with a generic visual processor emerging on the side) Dessì, Roberto Kharitonov, Eugene Baroni, Marco Aprenentatge automàtic Xarxes neuronals (Informàtica) Comunicació presentada a la 35th Conference on Neural Information Processing Systems (NeurIPS 2021) celebrada del 6 a 14 de desembre de 2021 de manera virtual. Inclou material suplementari: Appendix to Interpretable agent communication from scratch (with a generic visual processor emerging on the side) As deep networks begin to be deployed as autonomous agents, the issue of how they can communicate with each other becomes important. Here, we train two deep nets from scratch to perform large-scale referent identification through unsupervised emergent communication. We show that the partially interpretable emergent protocol allows the nets to successfully communicate even about object classes they did not see at training time. The visual representations induced as a by-product of our training regime, moreover, when re-used as generic visual features, show comparable quality to a recent self-supervised learning model. Our results provide concrete evidence of the viability of (interpretable) emergent deep net communication in a more realistic scenario than previously considered, as well as establishing an intriguing link between this field and self-supervised visual learning. 2022-03-04T06:49:20Z 2022-03-04T06:49:20Z 2021 info:eu-repo/semantics/conferenceObject info:eu-repo/semantics/publishedVersion Ranzato M, Beygelzimer A, Nguyen K, Liang PS, Vaughan JW, Dauphin Y, editors. Pre-proceedings of the 35th Conference on Neural Information Processing Systems (NeurIPS 2021); 2021 Dec 6-14. [S.l.]: NeurIPS; 2021. [13 p.] © Roberto Dessì, Eugene Kharitonov, Marco Baroni info:eu-repo/semantics/openAccess Neural Information Processing Systems (NeurIPS)