[#47] Network mutual information measures for graph similarity & Re-visiting voluntary sacrifices in academia
In network analysis, tasks like clustering or detecting anomalies often need a way to measure how similar two networks are. For this to work well, the similarity measure must be clear, reliable, and able to distinguish real patterns from random noise. In this work, we develop new graph similarity measures based on information theory principles. These measures help capture shared information between networks, even when networks are viewed at different levels of detail. We test our approach on both real and synthetic data and show that it effectively reveals key similarities in networks across various scales.
Two years ago, NetPlace hosted a talk about volunteer work in academia, exactly while I was volunteering for two academic events. At that time, nearly half of the respondents felt exploited. How have my perspective and the community’s views evolved over the past two years? (Find more details and the recording of the discussion here.)
In our last seminar, we had the pleasure of hosting Helcio Felipe, a PhD candidate from the Central European University, as well as running our first (unofficial) hybrid seminar, with in-person attendees in Vienna!
In the first part, Helcio introduced us to the problem of graph similarity that he and his collaborators are working on. He then presented possible tools to quantify and analyze graph similarity on micro, macro, and meso scales, using insights from information theory.
In the discussion that followed, Helcio shared his experience with volunteering in academia, leading to an audience discussion on how much time and effort is reasonable to dedicate to volunteering, as well as potential red flags to watch out for beforehand.