18485457. Scalable Self-Supervised Graph Clustering simplified abstract (GOOGLE LLC)

Scalable Self-Supervised Graph Clustering

Organization Name

GOOGLE LLC

Inventor(s)

Prateek Jain of Bangalore (IN)

Inderjit Singh Dhillon of Berkeley CA (US)

Fnu Devvrit of Austin TX (US)

Aditya Sinha of Champaign IL (US)

Scalable Self-Supervised Graph Clustering - A simplified explanation of the abstract

This abstract first appeared for US patent application 18485457 titled 'Scalable Self-Supervised Graph Clustering

Simplified Explanation

The method described in the patent application involves training a machine learning model using a graph structure and feature attributes to determine an encoded graph based on the model's output. The model includes a graph convolutional network layer and the encoded graph consists of nodes and paths connecting them. Positive samples are selected through random walks along the paths of the encoded graph, while negative samples are randomly sampled from the nodes. The model's learnable parameters are updated based on the loss value calculated during training.

• Machine learning model training method:

   - Utilizes graph structure and feature attributes
   - Determines an encoded graph using a graph convolutional network layer
   - Selects positive samples through random walks and negative samples by random sampling
   - Updates learnable parameters based on loss value

Potential Applications

The technology can be applied in: - Social network analysis - Recommendation systems - Bioinformatics

Problems Solved

- Efficiently training machine learning models on graph data - Handling graph structures with complex relationships

Benefits

- Improved accuracy in predictions - Better understanding of graph data relationships

Potential Commercial Applications

Optimized for: - Social media platforms - E-commerce websites - Healthcare systems

Possible Prior Art

One potential prior art for this technology could be the use of graph convolutional networks in machine learning models for graph data analysis.

Unanswered Questions

How does the method handle large-scale graph structures efficiently?

The patent abstract does not provide details on how the method addresses scalability issues when dealing with massive graph structures.

What types of feature attributes are most effective in improving model performance?

The abstract does not specify the types of feature attributes that yield the best results when training the machine learning model.

Original Abstract Submitted

A method of training a machine learning model includes receiving training data comprising a graph structure and one or more feature attributes and determining an encoded graph based on applying the machine learning model to the graph structure and the one or more feature attributes. The machine learning model comprises a graph convolutional network layer. The encoded graph comprises one or more nodes and one or more paths connecting the one or more nodes. The method also includes selecting a plurality of positive samples through random walks along the one or more paths of the encoded graph, selecting a plurality of negative samples from the encoded graph by randomly sampling the one or more nodes of the encoded graph, determining a loss value, and updating, based on the loss value, one or more learnable parameter values of the machine learning model.