backwardtuning

Recurrent network weight tuning

This only an alpha version, better contact us to discuss, what it is and how to use it :)

We propose here a re-implementation of backward tuning as developped and descrined in (vieville et al 2017) allowing to adjust the parameters of a vectorial recurrent discrete-time dynamical system.

Package repository

• Package files: https://gitlab.inria.fr/line/aide-group/backwardtuning
• Package documentation: https://line.gitlabpages.inria.fr/aide-group/backwardtuning
• Source files: https://gitlab.inria.fr/line/aide-group/backwardtuning/-/tree/master/src
• Version 0.1.0
• License CECILL-C

Installation

User simple installation

• npm install git+https://gitlab.inria.fr/line/aide-group/backwardtuning.git

Co-developper installation

• See the related documentation

Please refer to the installation guide for installation.

Usage

npm script usage

npm install --quiet : installs all package dependencies and sources.
npm run build: builds the different compiled, documentation and test files.
npm test     : runs functional and non-regression tests.
npm run clean: cleans installation files.

Dependencies

• aidesys: Basic system C/C++ interface routines to ease multi-language middleware integration
• wjson: Implements a JavaScript JSON weak-syntax reader and writer

devDependencies

• aidebuild: Builds multi-language compilation packages and related documentation.

Author

• Thierry Viéville thierry.vieville@inria.fr