****Events

###### Benjamin Fish

When is it possible to infer network structure from data? In the first part of the talk, we investigate how to reconstruct social networks from voting data. In particular, given a voting model that considers social network structure, we aim to find the network that best explains the agents’ votes. For two models of voting, we give algorithms and lower bounds, characterizing cases where network recovery is possible and where it is computationally difficult. Despite the similarity of the two models, we show that their respective network recovery problems substantially differ in complexity and produce very different networks, and then show the results on real data. In the second part of the talk, we discuss how to learn the time scale of a dynamic network. As a pre-processing step before data analysis, typically you need to choose a bin size, where all edges occurring within a bin are aggregated together to form a sequence of graphs. In previous work, finding a bin size is often accomplished with an unsupervised heuristic, which does not take into account that each task an analyst wants to perform on the data requires a different bin size. Given this, we introduce new windowing algorithms that automatically adapt to the task the analyst wants to perform by treating windowing as a hyperparameter for the task. We show this approach outperforms previous approaches on real data.

## Want to be notified about upcoming NetSI events?

Sign up for our email list below!

Thank you! You have been added to our email list.

Oops! Something went wrong while submitting the form

When is it possible to infer network structure from data? In the first part of the talk, we investigate how to reconstruct social networks from voting data. In particular, given a voting model that considers social network structure, we aim to find the network that best explains the agents’ votes. For two models of voting, we give algorithms and lower bounds, characterizing cases where network recovery is possible and where it is computationally difficult. Despite the similarity of the two models, we show that their respective network recovery problems substantially differ in complexity and produce very different networks, and then show the results on real data. In the second part of the talk, we discuss how to learn the time scale of a dynamic network. As a pre-processing step before data analysis, typically you need to choose a bin size, where all edges occurring within a bin are aggregated together to form a sequence of graphs. In previous work, finding a bin size is often accomplished with an unsupervised heuristic, which does not take into account that each task an analyst wants to perform on the data requires a different bin size. Given this, we introduce new windowing algorithms that automatically adapt to the task the analyst wants to perform by treating windowing as a hyperparameter for the task. We show this approach outperforms previous approaches on real data.