OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Neural Graphs

Neural Graphs is an open-source Python research implementation of graph neural networks for learning equivariant representations of neural networks, accounting for permutation symmetry in network structures and parameter spaces.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Neural Graphs — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need research code and experiment instructions for studying permutation-equivariant representations of neural networks, rather than a hosted model or general-purpose training interface.

Who it is for

  • Researchers studying graph neural networks and representations of neural-network parameters.
  • Developers reproducing the repository's INR classification, INR style-editing or CNN-generalization experiments.

Documented uses

  • Explore a graph-based approach to representing neural-network structures and weights.
  • Reproduce INR classification experiments using MNIST or Fashion MNIST INR data.
  • Study INR style editing with MNIST-derived representations.
  • Investigate CNN generalization using grayscale CIFAR-10 and CNN Wild Park.
Before you choose

Requirements & limits

The repository is public and specifies the MIT license.

  • The documented environment is based on Python 3.9, PyTorch 2.0.1 and PyG 2.3.0, with CUDA 11.7 listed in the install command.
  • Experiment-specific data downloads and reproduction steps are in the individual experiment folders.
  • The README links an introduction notebook for INR classification.
  • Learning to optimize is listed as coming soon, not as a documented ready-to-run experiment.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the research environment

Create the documented Conda environment and install the Neural Graphs repository in editable mode.

Before you start

  • Conda
  • Python 3.9
  • PyTorch 2.0.1, torchvision, torchaudio and PyTorch CUDA 11.7
  • PyG 2.3.0, pytorch-scatter, hydra-core, einops and opencv-python

Steps

  1. Create and activate a clean Conda environment named neural-graphs with Python 3.9.

  2. Install the specified PyTorch, vision/audio, CUDA and PyG packages using the README's Conda commands.

  3. Install hydra-core, einops and opencv-python with pip.

  4. Clone the repository, enter its directory and run pip install -e . as shown in the README.

Expected outcome

The repository is installed in editable mode in the documented research environment.

Guide limitations

  • The README's clone command contains a duplicated https:// in the repository URL; verify the clone URL before using it.
  • This setup guide does not run an experiment. Each experiment folder supplies separate data and reproduction instructions.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
mkofinas/neural-graphs · Official homepage title · retained in its source language
How we review sourcesSuggest a correction