Authors: Md Shamim Hussain, Mohammed J. Zaki, Dharmashankar Subramanian

Published on: February 07, 2024

Impact Score: 8.3

Arxiv code: Arxiv:2402.04538

Summary

• What is new: Introduction of the Triplet Graph Transformer (TGT) that enhances direct communication between neighboring pairs in graphs, along with a novel three-stage training procedure and stochastic inference.
• Why this is important: Graph transformers lack efficient pair-to-pair communication for tasks like molecular property prediction.
• What the research proposes: TGT enables direct neighbor pair communication in graphs using triplet attention and aggregation mechanisms, improving the prediction of interatomic distances and molecular properties.
• Results: Achieved new state-of-the-art results on benchmarks PCQM4Mv2, OC20 IS2RE, QM9, MOLPCBA, LIT-PCBA, and demonstrated generality on the traveling salesman problem (TSP).

Technical Details

Technological frameworks used: Triplet Graph Transformer (TGT)

Models used: Triple attention and aggregation mechanisms, stochastic inference

Data used: 2D graphs for molecular property prediction

Potential Impact

Chemical and pharmaceutical industries, logistics and optimization solution providers.

Want to implement this idea in a business?

We have generated a startup concept here: MolIntel.