Skip to main content

Gaussian Processes for orientation preference maps

Project description

Gaussian Processes for Orientation Preference Maps

This package is a Python implementation of a Gaussian Process (GP) method for inferring cortical maps (Macke et al., 2011), based on their original MATLAB-code. Compared to conventional (vector averaging) approaches, the method computes better maps from little data and can be used to quantify the uncertainty in an estimated orientation preference map (OPM). It also provides a principled way to choose the parameters in the smoothing step and incorporate inhomogeneous noise variances as well as noise correlations.

The basic idea is to specify a prior over OPMs as a GP. This means that a-priori assumptions about correlations of the pixels in a map are specified as a covariance function K. In this case, we use a Difference of Gaussians (DoG) kernel. This covariance function encodes our assumptions that OPMs are smooth and have a semi-periodic structure. This prior is then updated via Bayes' rule with imaging data to obtain a smooth posterior estimate of an OPM.

My notes on the method (including derivations not in the original paper) can be found here.

Installation & Usage

You can clone this package and install all its dependencies using

git clone
python -m pip install -r requirements.txt

Alternatively, this module is now pip-installable via

python -m pip install gp-maps-python

A number of Jupyter notebooks illustrate how to use the package:


contains basic functionality for handling OPMs:

  • sample from this prior to obtain a ground truth OPM
  • visualize OPMs
  • simulate neural responses given an OPM and a set of stimuli
  • compute maximum likelihood/vector averaging estimate of the OPM
  • smooth the OPM with a Gaussian kernel


shows details about the GP method:

  • covariance function (DoG kernel)
  • estimate hyperparameters of the prior covariance function from the empirical map
  • learn noise model iteratively from data
  • posterior inference


serves as an illustration on how to use GPs to infer OPMs from real imaging data (replicating the results of Macke et al., 2011):

  • visualize inhomogeneous and correlated noise structure of the data
  • obtain smooth estimate based on "ground truth"
  • GP posterior inference on real data

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

gp-maps-python-0.0.3.tar.gz (18.2 kB view hashes)

Uploaded Source

Built Distribution

gp_maps_python-0.0.3-py3-none-any.whl (38.4 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page