Accenture/AmpliGraph

What it solves

AmpliGraph is designed to predict unseen or missing links between concepts in a knowledge graph. It helps users discover new knowledge, complete large knowledge graphs with missing statements, and generate embeddings for the entities within those graphs.

How it works

The library uses neural machine learning models to create knowledge graph embeddings—vector representations of concepts in a metric space. It then applies model-specific scoring functions to these embeddings to predict novel links between entities.

Who it’s for

It is intended for developers and researchers who need to perform relational learning on knowledge graphs, evaluate new relational models, or perform knowledge discovery tasks like clustering entities and predicting duplicates.

Highlights

• Keras-style APIs: Simplifies the process of learning models that predict links in knowledge graphs.
• GPU-Ready: Built on TensorFlow 2, allowing for seamless training acceleration on both CPU and GPU devices.
• Extensible: Users can create their own knowledge graph embedding models by extending the base estimators.
• Multiple Model Support: Includes implementations of models such as TransE, DistMult, ComplEx, HolE, and RotatE.