Marta Tatu

Title: Marta Tatu: Inventor in Dynamic Graph Pattern Discovery.

Introduction

Marta Tatu is an inventor based in Murphy, Texas. Her patent work includes a method for discovering patterns in continuous-time dynamic graphs through graph neural network embeddings.

Latest Patents

Tatu is associated with one patent, titled “Unsupervised Pattern Discovery Using Dynamic Graph Embeddings.” The invention describes receiving source and destination node embeddings from a graph neural network, clustering the destination embeddings, and removing embeddings assigned to a noise group. It then clusters the remaining signal embeddings and uses the resulting groups, together with the source and destination embeddings, to identify patterns in the graph.

This approach is designed to help uncover meaningful structure in dynamic graph data without relying on labeled examples. Tatu’s patent portfolio includes 1 patent.

Career Highlights

Tatu is associated with Raytheon Company. Her patented work focuses on graph-based pattern discovery, combining neural network embeddings with clustering to distinguish signal from noise in continuously changing data.

Collaborations

Joshua A. Binkley is identified as Tatu’s coworker. The available information does not specify his role in the patent or the invention.

Conclusion

Marta Tatu’s patent highlights an approach to unsupervised pattern discovery that uses dynamic graph embeddings and clustering to identify meaningful relationships in evolving data.