lytemaps 0.0.3

Lightweight implementation of neuromaps

lytemaps

This repository contains an implementation of a minimal fork of neuromaps functionality. If you're an end user, this is likely not for you -- this implementation exists mostly for hypercoil developer purposes. As developers, we'd like to use certain neuromaps operations without burdening our software with the bulky dependencies (nilearn, sklearn and Connectome Workbench) that the full install of neuromaps requires. This implementation is intended to support the subset of functionality that requires only core Scientific Python packages together with the essential nibabel and pooch for handling data fetch operations. (But we're not there yet -- in particular, many transforms still require Workbench at this time.) Obviously, functionality here is limited -- inter alia, this means that null models aren't included for now. As development progresses, we'll index the neuromaps functions that are implemented below.

If for some reason you still decide to use this repository, please follow the citation prescriptions from Neuromaps and pooch. In particular, specify that you used lytemaps in your methods section, specify that lytemaps comprises code from neuromaps with a pooch-based downloader backend, and cite the neuromaps and pooch papers.

License

Most code in this repository is taken directly from the neuromaps repository, and is therefore licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License. The exception to this is several small utility functions in src/lytemaps/utils.py and src/lytemaps/datasets/utils.py, which are taken from the nilearn and sklearn repositories and are therefore licensed under the 3-clause BSD license. As we redevelop code into our own implementations, we will relicense. See the LICENSE file for more details.

Roadmap

• Implement pooch-based data fetchers for neuromaps datasets
• Add support for token-based authentication for OSF datasets
• Add support for querying the templateflow API
• Use nitransforms wherever possible for surface-to-surface, volume-to-surface, and surface-to-volume transforms
• Use neuroimaging tensor library backend for operations originally implemented with nilearn and sklearn