This repository contains the code and files that are associated with the paper titled "Chance, long tails, and inference in a non-Gaussian, Bayesian theory of vocal learning in songbirds".
Folder "code":
generate_figures.m
generates all the panels shown in Fig. 2, Fig. 3, Fig. 4, and Fig. 5 of the paper.
dynamical_Bayesian.m
simulates one iteration of the dynamical Bayesian filter model shown in Fig. 1 of the paper.
songbird_single_timescale_stable.m
simulates the four constant-shift experiments, and the staircase-shift experiment.
song_acquisition.m
simulates the song acquisition period during which the bird begins to sing and gradually refines its song until the beginning of the adulthood.
shift_experiment.m
simulates the shift experiment given the shift size for each day.
generate_kernels_likelihoods.m
generates one kernel, and two likelihoods that are used in the Bayesian filter given a set of input parameters.
get_shifted_likelihood.m
generates shifted likelihood corresponding to the shifted auditory channel.
stable_distri_laguerre_bergstrom.m
simulates the 1d symmetric stable distribution.
powerlaw_distribution.m
simulates the 1d symmetric power law distribution.
histnorm.m
is some helper function adapted from Arturo Serrano.
test_code.m
is for authors' own usage and can be ignored. But one can use it to test how those learning curves change when parameters change.
Folder "data":
In this folder, processed experimental data are stored.
Folder "figures":
The Fig. 2, Fig. 3, Fig. 4, and Fig. 5 in the paper, which are generated from generate_figures.m
, are stored in this folder.
Folder "variables":
In this folder, theoretical variables simulated from the model and optimization are stored.
Folder "stable_distribution_compare":
Different methods to simulate the stable distribution are compared in this folder.