LDA Network Performance Model for Whole-Node Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network performance models, such as the Gaussian mixture model, are unreliable for representing the performance of the whole network as they only account for the nodes providing the performance data and do not accurately reflect the performance of other network nodes with similar aggregation features.
Innovation Solution
A method and apparatus using Latent Dirichlet Allocation (LDA) to establish a network performance model by determining parameters α and β, which represent the correlation and distribution pattern of performance data across different network nodes, enabling a reliable model that fits the whole network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a Gaussian mixture model is used to establish the network performance model, then the model can be established using sample performance data from network nodes, but the model is not reliable for showing the performance of the whole network and only fits the nodes that provided the data
Solution Approach 1:
The patent transforms the network performance modeling approach by changing the mathematical parameters from simple Gaussian distributions to Latent Dirichlet Allocation (LDA) model parameters. This involves introducing topic distribution parameters and document-topic distribution parameters that capture the underlying structure of network performance data across different nodes, enabling the model to generalize from sample nodes to the entire network.
Solution Approach 2:
The patent creates a virtual copy of the network performance characteristics by using LDA to infer the performance distribution of non-sampling nodes based on the sampling nodes. The topic models capture the essential performance patterns and apply them to represent the whole network, effectively copying the performance characteristics from observed nodes to unobserved nodes.
2Ease of operation
If performance data is collected from only a part of network nodes, then the data collection process is simplified, but the model cannot accurately represent the performance of other network nodes with similar aggregation features
Solution Approach 1:
The LDA-based network performance model achieves universality by creating a generalized representation that applies to all network nodes with similar aggregation features. The topic distribution parameters learned from sampling nodes serve as universal characteristics that can represent the performance of any node within the same aggregation group, making the model universally applicable across the network.
Solution Approach 2:
The patent introduces topic distributions as an intermediary layer between the sampled performance data and the inference of non-sampled node performance. The LDA model acts as a mediator that translates observed performance characteristics into latent topic structures, which then serve as the basis for inferring the performance of unobserved nodes, bridging the gap between limited samples and comprehensive network representation.
Data Source
AI summary
A method and apparatus for establishing a network performance model. The method includes: determining, according to performance data provided by network nodes and the probability of the performance data, a parameter α showing the correlation of the performance data of different network nodes in a whole network and a parameter β showing the distribution pattern of the performance data in the network; and establishing a Latent Dirichlet Allocation, LDA, network performance model by using the determined parameter α and the parameter β.