Skip to content

Latest commit

 

History

History
100 lines (76 loc) · 7.91 KB

README.md

File metadata and controls

100 lines (76 loc) · 7.91 KB
Raw

Neural Synchrony for Figure Ground Segregation

Analysis and simulation code accompanying the article:

Karimian, A., Roberts, M.J., De Weerd, P., & Senden, M. (n.d.). Gamma Synchrony Mediates Figure Ground Perception. Manuscript submitted.

Abstract

Gamma synchrony is ubiquitous in visual cortex, but whether it contributes to perceptual grouping remains contentious based on observations that gamma frequency is not consistent across stimulus features and that gamma synchrony depends on distances between image elements. These stimulus dependencies have been argued to render synchrony among neural assemblies encoding components of the same object difficult. Alternatively, these dependencies may shape synchrony in meaningful ways. Using the theory of weakly coupled oscillators (TWCO), we demonstrate that stimulus dependence is crucial for gamma's role in perception. Synchronization among coupled oscillators depends on frequency dissimilarity and coupling strength, which in early visual cortex relate to local feature dissimilarity and physical distance, respectively. We manipulated these factors in a texture segregation experiment wherein human observers identified the orientation of a figure defined by reduced contrast heterogeneity compared to the background. Human performance followed TWCO predictions both qualitatively and quantitatively, as formalized in a computational model. Moreover, we found that when enriched with a Hebbian learning rule, our model also predicted human learning effects. Increases in gamma synchrony due to perceptual learning predicted improvements in behavioral performance across sessions. This suggests that the stimulus- dependence of gamma synchrony is adaptable to the statistics of visual experiences, providing a viable neural grouping mechanism that can improve with visual experience. Together our results highlight the functional role of gamma synchrony in visual scene segmentation and provide a mechanistic explanation for its stimulus-dependent variability.

Repository Overview

Directory Structure

The project is organized into the following directories:

├── config
│   ├── analysis
│   ├── plotting
│   └── simulation
├── data
├── notebooks
├── results
├── scripts
│   ├── analysis
│   ├── info
│   ├── plotting
│   ├── simulation
│   └── statistics
└── src

config

This directory contains TOML configuration files for the analysis, plotting, and simulation scripts. These files specify the parameters used by the scripts and allow for easy configuration of the study.

Note that for reasons of reproducibility, the random seed is fixed in the simulation.toml config file. This ensures that the results of the simulations can be reproduced exactly.

data

This directory is an empty placeholder for the data collected and analyzed in the study. The Snakemake workflow of the simulation and analysis components of the study automatically downloads CSV data files from Zenodo and places them in this directory. The data includes human psychophysics data on figure ground segregation in texture stimuli.

Karimian, M., Mark, R., De Weerd, P., & Senden, M. (2024). Human Psychophysics Dataset on Figure Ground Segregation in Texture Stimuli [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10817187

notebooks

This directory contains Jupyter notebooks for exploring and reporting results as well as for generating additional figures and performing additional statistical analyses.

results

This directory contains the results generated by the scripts. Results are organized into subdirectories following the same structure as the organization of scripts into subdirectories.

scripts

This directory contains scripts for simulations, analysis, statistics, plotting, and querying system information. Each script is designed to perform a specific task within the Snakemake workflow. The scripts are organized into subdirectories based on their functionality.

src

This directory contains Python source code for the study. The init.py file is used to mark the directory as a Python package. The source code includes classes and functions for model simulations, analyzing data, statistics, and generating figures.

Workflow

The workflow of the project is defined in the Snakefile. The Snakefile specifies a set of rules that define how to generate the desired results from the input data. The workflow uses the Snakemake workflow management system to automatically execute the rules and generate the results.

The workflow consists of the following steps:

  1. Download the data from Zenodo.
  2. Run system information queries to gather information about the system used to run the workflow.
  3. Perform main statistical analyses on human psychophysics data using generalized estimating equations (GEE).
  4. Pre-process human psychophysics data to obtain behavioural Arnold tongues and estimate optimal psychometric parameters for each session.
  5. Evaluate assumption of local learning by analysing transfer session results.
  6. Explore the parameter space of the model to compare empirically derived parameters to optimal parameter range.
  7. Estimate the learning rate of the model using cross-validation.
  8. Simulate the learning experiment using the estimated learning rate.
  9. Test the model predictions against the human psychophysics data.
  10. Generate figures to visualize the results.

Running the Workflow

There are two ways to run the workflow:

  1. Locally: To run the workflow locally, navigate to the root directory of the project and execute the following command:
snakemake --cores [number_of_cores]

Replace [number_of_cores] with the desired number of CPU cores to be used for the workflow. This command will execute the workflow and generate the results in the results directory.

Note that model simulations involve parallel processes implemented with the multiprocessing tools. The config/simulation/simulation.toml file specifies the number of CPU cores that should be used. If the machine you run this workflow on has fewer cores than specified in the config file, the workflow will still run. However, all of your machine's cores will be used, and the number of blocks of the simulations will be adjusted to be an integer multiple of the number of used cores. This may lead to slight numerical deviations from results reported in the manuscript. It is recommended to adjust the number of cores in the config file and in the --cores flag if you would like to use fewer cores and to ensure that the number of simulation blocks match those used to produce results reported in the manuscript.

For example, if you want to run the workflow using 10 cores, execute the following command:

snakemake --cores 10

This will run the workflow using 10 CPU cores and generate the results in the results directory.

  1. Docker container: The workflow can also be run as a Docker container. To do this, build the Docker image using the following command:
docker build -t adaptive-synchronization .

Then, run the container using the following command:

docker run --rm -it -v $(pwd)/results:/workflow/results adaptive-synchronization snakemake --cores 30

This will run the workflow using 30 CPU cores and generate the results in the results directory.

Note that the -v flag maps the results directory from the host machine to the /workflow/results directory inside the container, allowing the workflow to write its results to the host machine.. The --rm flag removes the container after it has finished running.

The Dockerfile used to build the image is included in the root directory of the project. It is based on the latest Snakemake image and includes all the necessary dependencies for running the workflow. The CMD in the Dockerfile runs the Snakemake workflow with 30 cores by default. However, this can be overridden by specifying a different number of cores when running the container.