What is personalized analysis of graphs?
Personalized Recommendations Controlling the Influence of Popularity Bias @ tullys
It is well known that personalized recommendations tend to favor popular content for everyone. This phenomenon, known as popularity bias, is a significant issue that undermines the user-centricity of recommendation results and the fairness of the system. Therefore, this research investigates methods to universally control the impact of popularity bias by adjusting the scope of analysis and applying weighting based on content popularity.
Random Walk for Local Community Detection with Weighting Based on Connectivity @ esty
When personalizing graph analysis according to user interests, methods for detecting local communities around users—particularly those that explore neighborhoods via random walks—prove useful. Existing approaches uniformly add surrounding nodes to communities. In contrast, this research achieves community detection that better reflects user interests by performing weighted random walks that focus on the connectivity between adjacent nodes.
Fast Personalized Gragh Summarization by Random Walk Control @ sigma
By summarizing graphs according to user interests, we can reduce the spatial cost of graph analysis while reducing the impact on analysis accuracy. Current graph summarization methods require traversal of the entire original graph for each time based on user interest, which is time-consuming. This research makes personalized summarization faster by replacing part of the original graph with a pre-summarized graph instead of generating the entire summary graph, which allows some computation for graph summarization in advance.