This is a part of the code of my master's project "Using the stimulus specifc information to investigate the effect of neural correlations on sensory coding".
It was an interdisciplinary research project between computational neuroscience and information theory.
These 4 jupyter notebooks corresponds to 3 different parts of my project :
- Single Neuron Study
- 2-neurons System
- Small Population Study
- Retinal direction-selective cells data analysis
The way a population of sensory neurons encodes information from external stimuli has been investigated for decades, but the effect of response noise on the encoding capacity of neurons remains largely unknown. The classical view on the subject is supported by theoretical arguments based on the Fisher Information - a information theoretic metric that assumes unrealistic condition. The aim of the project is to challenge this classical view of neural noise by using the Stimuli Specific Information (SSI), a more recent and refined metric, to investigate the effect of neural noise on sensory encoding. We start by a theoretical analysis of single neurons and neural populations responses to stimuli, and build a population model where noise correlation accounts for the interactions between neurons. Thereafter, we use this theoretical grounding to analyse direction-selective cells recordings through the stimulus specific information, by fitting our model to a sub-population of retina cells. This project led to interesting results on single neuron encoding and to conjectures on the effect of different noise characteristics on the ability of neuron population to encode sensory inputs. Ultimately, this project raises more questions that it solves, and could be a fertile ground for further analysis of sensory neurons
