Modeling the role of night/day alternations in the development of the zebra finch song
Project: Modeling the role of night/day alternations in the development of the zebra finch song
Laboratory: Institut des Systèmes Intelligents et de Robotique (ISIR, UMR7222)
Affiliation: CNRS / Sorbonne Université
Address: 4 place Jussieu, CC173
Name: Raja Chatila
Phone number: 0144275197
Name: Benoît Girard
Phone number: 0144276381
Subject Keywords: development of vocalizations, reinforcement learning, evolutionary computation, zebra finch
Summary of lab’s interests: computational neuroscience, learning, decision-making, basal ganglia, developmental robotics
Project summary: The Zebra Finch is an animal model of language acquisition. It learns a song from a male tutor during a very specific period (between 30 and 90 days after hatching), after which it will be crystallized. It has been shown that this learning has a very specific trajectory with regards to day/night periods (Derégnaucourt, S., Mitra, P. P., Fehér, O., Pytte, C., & Tchernichovski, O. (2005). How sleep affects the developmental learning of bird song. Nature, 433(7027), 710-716.): while the quality of the reproduction of the tutor song increases during the day, it declines during the night. However this night degradation seems to have a functional role in learning, as those birds who have the largest night degradations produce the best imitations of the tutor’s song at the end of the learning period.
The goal of this project is to test hypotheses on the learning mechanisms operating during the day and the night, so as to explain this apparently paradoxical result. For this, we will use technics from the reinforcement learning and the evolutionary computations domains. The evaluation of the model will be performed using experimetal data gathered by Sebastien Derégnaucourt (from the LECD lab).
The PhD will be prepared at ISIR Lab (CNRS & SU, Paris) under the supervision of Benoît Girard and Stéphane Doncieux, and in direct collaboration with Sebastien Derégnaucourt (LECD, Univ. Paris Ouest). It follows a preliminary work done in collaboration between ISIR and LECD on this topic that has resulted in a first model presented at the European Birdsong meeting (Ecoffet, P. and Girard, B. and Doncieux, S., 2017. Computational model of birdsong learning and influence of sleep and replays with Zebra Finches. European Birdsong meeting (EBM). Bordeaux, France).
Interdisciplinary aspect of the project: The project will be composed of modeling activities carried out at ISIR, but in strong collaboration with the LECD. The model will be tested on the data already acquired at LECD, requiring deep analysis of this data, and new experiements explicitely testing predictions of the model will be carried out at LECD.