17806075. TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH STRUCTURAL INFORMATION ENCODING simplified abstract (Microsoft Technology Licensing, LLC)
TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH STRUCTURAL INFORMATION ENCODING
Organization Name
Microsoft Technology Licensing, LLC
Inventor(s)
TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH STRUCTURAL INFORMATION ENCODING - A simplified explanation of the abstract
This abstract first appeared for US patent application 17806075 titled 'TRANSFORMER-BASED GRAPH NEURAL NETWORK TRAINED WITH STRUCTURAL INFORMATION ENCODING
Simplified Explanation
The abstract describes a computing system that uses a processor to train a model for predicting energy changes in molecular systems. The system provides a training data set that includes pre-transformation molecular graphs and corresponding energy parameter values. The pre-transformation graphs consist of normal nodes connected by edges representing distances and bonds between the nodes.
The processor is configured to encode structural information in the pre-transformation graphs as learnable embeddings. This structural information describes the relative positions of the atoms represented by the normal nodes. It includes an edge encoding that represents the type of bond between pairs of normal nodes and a spatial encoding that represents the shortest path distance along the edges between pairs of normal nodes.
- The computing system trains a model to predict energy changes in molecular systems.
- The training data set includes pre-transformation molecular graphs and energy parameter values.
- The pre-transformation graphs consist of normal nodes connected by edges representing distances and bonds.
- Structural information in the graphs is encoded as learnable embeddings.
- The structural information includes edge encoding and spatial encoding.
- The edge encoding represents the type of bond between pairs of normal nodes.
- The spatial encoding represents the shortest path distance between pairs of normal nodes.
Potential Applications
- Predicting energy changes in molecular systems.
- Drug discovery and development.
- Chemical reaction optimization.
- Material design and optimization.
Problems Solved
- Accurate prediction of energy changes in molecular systems.
- Efficient encoding of structural information in molecular graphs.
- Improved understanding of the relative positions of atoms in a molecular system.
Benefits
- More accurate predictions of energy changes in molecular systems.
- Faster and more efficient training of models.
- Improved understanding of molecular structures and properties.
- Potential for advancements in drug discovery, material design, and chemical reaction optimization.
Original Abstract Submitted
A computing system is provided, including a processor configured to, during a training phase, provide a training data set, including a pre-transformation molecular graph and post-transformation energy parameter value representing an energy change in a molecular system following an energy transformation. The pre-transformation graph includes a plurality of normal nodes connected by edges representing a distance and a bond between a pair of the normal nodes. The processor is further configured to encode structural information in each pre-transformation molecular graph as learnable embeddings, the structural information describing the relative positions of the atoms represented by the normal nodes. The structural information includes an edge encoding representing a type of bond between a pair of normal nodes in each pre-transformation molecular graph, and a spatial encoding representing a shortest path distance along the edges between a pair of normal nodes in each pre-transformation molecular graph.
