Welcome to tart!

🔥 tensor abstractions for fast analysis of graph relations 🔥

pip install tart-lib

tart is a library for learning approximate and abstract relations between graphs. It is currently designed for a very general graph relation – the subgraph relation (⊆). tart learns tensor representations that encodes the subgraph relation between graphs and provides various APIs to train, evaluate, and predict over these representations.

tart is:

1. 🌐 General Purpose (json graphs and type-agnostic features)

2. 🪢 Extensible (supports custom models and encoders)

3. 🔥 Fast (tensor comparisons for relation analysis)

tart in 3 steps:

Load JSON Graphs

Specify Configs

Use APIs

Note

This project is under active development. Please file an issue or ask the contributors if you encounter bugs.

Contents

Overview

• Installation
  • Install tart from PyPI
  • Install tart from GitHub
• Architecture

Tutorials

• Learning using tart
  • Step 1: Creating a dataset
  • Step 2: Specify configs
  • Step 3: Learn a tart model
• Prediction using tart
  • Task 1: Subgraph Matching
  • Task 2: Subgraph Counting

Package documentation

• API documentation