SKIS Hyper-Graph Sketching for Influence Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current diffusion analysis techniques face challenges in computational efficiency and scalability when dealing with massive-scale online social networks, leading to deficiencies in accuracy and memory efficiency for influence maximization and estimation.
Innovation Solution
The Importance Influence Sampling Sketch (SKIS) methodology employs a sampling scheme that randomly evaluates subsets of nodes based on probability, generating sketches with small variation and high concentration, while reducing memory footprint and processing time, and includes a hybrid technique for estimating local influence to improve estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional diffusion analysis techniques are used on massive-scale networks, then influence estimation can be performed, but computational efficiency and scalability deteriorate significantly
Solution Approach 1:
The patent segments the massive network into multiple smaller subgraphs through random sampling. Instead of analyzing the entire network at once, it divides the problem into manageable pieces by selecting random subsets of nodes and edges, which can be processed independently and then aggregated to estimate overall influence metrics.
Solution Approach 2:
The patent extracts a representative sample from the full network by randomly selecting nodes and edges according to specific probability distributions. This extraction creates a simplified version of the network that retains the essential structural properties needed for influence estimation while being computationally tractable.
2Quantity of substance
If traditional sketching techniques are used, then memory space is reduced, but accuracy of influence estimation deteriorates
Solution Approach 1:
The patent applies local quality by using different sampling probabilities for different parts of the network. Instead of uniform sampling, it assigns higher probabilities to nodes and edges that are more likely to be influential, creating a non-uniform distribution that prioritizes important regions while still maintaining a compact representation.
Solution Approach 2:
The patent changes the sampling parameters dynamically based on network properties. It adjusts the number of samples, sampling probabilities, and subgraph sizes according to the specific characteristics of the network being analyzed, allowing optimization of the trade-off between memory usage and estimation accuracy for different scenarios.
3Measurement precision
If more samples are collected to improve estimation accuracy, then processing time increases significantly
Solution Approach 1:
The patent performs preliminary actions by pre-computing certain network properties and statistics before the actual influence estimation. It calculates degree distributions, identifies high-degree nodes, and prepares sampling probabilities in advance, which speeds up the main estimation process and reduces the computational burden during query time.
Data Source
AI summary
A hyper-graph sketching framework receives graph data set to generate non-singular reverse cascades in a network to assess influence dynamics therein. A high-quality influence oracle for seed sets uses the resulting sketch data set generated from the non-singular reverse cascades to output estimation of influence.