Virtual Link Extension Network for Stochastic Clustering Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In clustering methods using stochastic processes, non-clique network structures lead to longer calculation times and false steady states due to the difficulty in probability diffusion and mixing, particularly when nodes are not directly linked.
Innovation Solution
An information processing apparatus that generates an extension network by adding virtual links between unconnected nodes, allowing for faster convergence of stochastic processes by creating shortcuts for probability transmission, thereby facilitating quicker mixing of probabilities within the network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If iterative calculations of stochastic process are performed on a non-clique network, then clustering operation is performed, but calculation time becomes excessively long and convergence to steady state is delayed
Solution Approach 1:
The invention performs preliminary actions by adding virtual links to create extension networks before the iterative stochastic process calculations begin. These virtual links pre-establish shortcut paths that will accelerate probability diffusion during the subsequent clustering calculations, thereby reducing calculation time while maintaining clustering accuracy.
Solution Approach 2:
Virtual links serve as intermediaries between nodes that are not directly connected by real links. These intermediary virtual links provide alternative pathways for probability flow, enabling faster diffusion and mixing of probabilities across the network without altering the fundamental cluster structure, thus resolving the contradiction between accurate clustering and fast convergence.
2Productivity
If virtual links are added to create extension network, then probability transmission is accelerated and convergence is improved, but network structure becomes more complex
Solution Approach 1:
The invention changes the network structure parameter by adding virtual links with specific properties (teleportation capability) to the original network. This parameter change creates extension networks that maintain the original network's essential clustering characteristics while introducing new pathways that accelerate probability diffusion, thereby improving convergence speed without fundamentally altering the network's cluster topology.
Data Source
AI summary
An information processing apparatus includes an acquisition unit that acquires information related to a network including a group of nodes, and a group of real links that connects the nodes, an adding unit that generates an extension network by adding to the network at least a virtual link that link nodes that are not linked by a real link, and a clustering unit that clusters the group of the nodes by iteratively calculating a stochastic process of a transition between the nodes linked by the real link and the virtual link in the extension network until the stochastic process is regarded to be in a steady state.