Skip to content

Master MVA - Computational Neuroscience project

Notifications You must be signed in to change notification settings

d-dawg78/MVA_Neuro

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 

Repository files navigation

Final Project Submission for the Computational Neuroscience course at the École Normale Supérieure (Master MVA)!

Group: Dorian Desblancs

Abstract:

In this project, we explore the field of efficient coding. We focus on the efficient coding of sounds using independent component analysis (ICA). The methodology used was heavily inspired by Michael S. Lewicki’s seminal paper Efficient Coding of Natural Sounds [16]. We used four sets of sounds, ordered from least to most periodic: environmental, speech, bird, and synthesizer. Our results demonstrate that the more periodic or harmonic a set of sounds is, the more sinusoidal its sources are. We notice that more random, environmental noises are split into components resembling wavelets. On the contrary, harmonic bird sounds are split into components resembling sinusoids. This work ties into the conclusions drawn by Lewicki. The center frequencies and bandwidths obtained by the set of derived auditory filters are similar to those found in the cochlear nervous system. This further reinforces the idea that information theory is linked to the brain, most notably in the realm of sparse coding.