Method and system for extracting information from graphs
A computer implemented method for classifying molecular structures is provided. The method includes representing the elements and atoms in a molecular structure as nodes and the bonds as relations as a first graphical representation. The method also includes mapping said first graphical representation into a second graphical representation wherein the mapping comprises using an attention mechanism, said attention mechanism establishes the importance of specific feature vectors dependent on their neighbourhood and the relations between the feature vectors, said mapping transforming the feature vectors of the first graphical representation to transformed feature vectors in the second graphical representation. The method also includes combining the transformed feature vectors to obtain a third combined representation. The method also includes mapping said third combined representation to a feature vector indicating properties of the molecular structure.
1. A computer implemented method for classifying molecular structures, said method comprising:
representing elements and atoms in a molecule as nodes and bonds in a molecular structure as relations as a first graphical representation, the first graphical representation having a plurality of different types of relations between the nodes, the nodes comprising feature vectors with features of the atomic properties;
mapping said first graphical representation into a second graphical representation wherein the mapping comprises using an attention mechanism, said attention mechanism establishes the importance of specific feature vectors dependent on their neighbourhood feature vectors and the different types of relations between the feature vectors, said mapping transforming the feature vectors of the first graphical representation to transformed feature vectors in the second graphical representation;
combining the transformed feature vectors to obtain a third combined representation; and
mapping said third combined representation to a feature vector indicating properties of the molecular structure.
2. The method of claim 1 , wherein the attention mechanism is enacted by attention coefficients and there is a coupling between attention coefficients across different relations.
3. The method of claim 1 , wherein the attention mechanism is enacted by attention coefficients and the attention coefficients across a neighbourhood are normalised.
4. The method of claim 1 , wherein the attention mechanism is enacted by attention coefficients and the attention coefficients across a neighbourhood of nodes for one relation are normalised.
5. The method of claim 1 , wherein the nodes of the first graphical data structure representation each comprise a feature vector and the nodes of the second representation also each comprise a feature vector, the mapping converting the features of the feature vectors of the first graphical data structure representation into transformed feature vectors of the second graphical data structure representation.
6. The method of claim 5 , wherein the feature vectors of the first graphical data structure representation are transformed into feature vectors of the second graphical data structure representation via a linear transformation and an attention mechanism.
7. The method of claim 6 , wherein the attention mechanism is configured such that after the linear transformation, the relation between the linearly transformed feature vectors are independent of other relations.
8. The method of claim 7 , wherein the attention mechanism comprises attention coefficients derived from a product of a relation dependent vector with a concatenation of two related nodes where the feature vectors from the two related nodes have been transformed via said linear transformation.
9. The method of claim 8 , wherein the attention coefficients are derived from the said product using an activation function that allows normalisation over the neighbourhood of nodes.
10. The method of claim 6 , wherein matrices that define the linear transformation and the attention mechanism are decomposed.
11. The method of claim 1 , wherein the attention mechanism is a multi-head attention mechanism.
12. A non-transitory computer-readable medium comprising instructions, which when executed on a computer cause the computer to perform the method of claim 1 .
13. A system for classifying molecular structures, said system comprising a processor and a memory, the processor being configured to:
represent elements and atoms in a molecule as nodes and bonds in a molecular structure as relations as a first graphical representation, the first graphical representation having a plurality of different types of relations between the nodes, the nodes comprising feature vectors with features of the atomic properties;
retrieve from memory a pre-trained model adapted to map said first graphical representation into a second graphical representation wherein the pre-trained model uses an attention mechanism, said attention mechanism establishes the importance of specific feature vectors dependent on their neighbourhood feature vectors and the different types of relations between the feature vectors, said mapping transforming the feature vectors of the first graphical representation to transformed feature vectors in the second graphical representation;
combine the transformed feature vectors to obtain a third combined representation; and
map said third combined representation to a feature vector indicating properties of the molecular structure.
14. The system of claim 13 , wherein the processor comprises a GPU.