Ordered Data Matrix for Temporal Graph Neural Network Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep Tensor technologies face challenges in handling sequence information such as anteroposterior relationships in time due to arbitrariness in node sequence processing, leading to incorrect learning and erroneous detection when correlations between sensors change over time.
Innovation Solution
A learning device generates an ordered data matrix by adding order information to input data, performing tensor decomposition, and inputting it to a neural network for deep machine learning, allowing the learning of relationships that include temporal ordering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If graph structure data is input to Deep Tensor for abnormality detection, then the relationship among multiple sensors can be learned, but sequence information such as temporal order cannot be handled correctly
Solution Approach 1:
The patent adds a time dimension to the graph structure data, transforming it from a static graph to a temporal graph. This is achieved by creating multiple graph representations at different time points and stacking them along the time dimension, allowing the model to process both spatial relationships among sensors and temporal sequences of events.
Solution Approach 2:
The patent performs preliminary processing of graph structure data by pre-calculating graph features, adjacency matrices, and node embeddings before feeding them to the neural network. This preprocessing step organizes the data in a format that preserves temporal order and facilitates efficient processing during the learning phase.
2Ease of operation
If node sequence is processed arbitrarily in graph structure learning, then processing flexibility is improved, but temporal order information is lost
Solution Approach 1:
The patent introduces dynamic graph neural network operations that adaptively process graph structures at each time step. The model dynamically updates node embeddings and adjacency matrices based on temporal evolution, allowing flexible processing while preserving temporal order through sequential processing of time-stamped data.
Solution Approach 2:
The patent processes graph data in periodic time steps, creating a sequence of graph representations that capture temporal patterns. By dividing the continuous time stream into discrete periodic intervals and processing each interval in sequence, the model maintains temporal order while preserving processing flexibility.
Data Source
AI summary
A non-transitory computer-readable recording medium stores a learning program that causes a computer to execute a process including: extracting, from a plurality of data groups, a plurality of first data groups having an order; generating, for each data element corresponding to each of the first data groups, an ordered data matrix in which data elements having same order have value corresponding to relationship among the data elements and data elements having different orders have values corresponding to the different orders; and obtaining input tensor data by performing tensor decomposition with the ordered data matrix, inputting the input tensor data to a neural network at time of performing deep machine learning, performing deep machine learning of the neural network, and learning about method for the tensor decomposition.