Last month was a little slower, with only (unless we missed some) three papers: two papers appearing, and one that was overlooked from the month before.
Stability of Homomorphisms, Coverings and Cocycles I: Equivalence, by Michael Chapmanand Alexander Lubotzky (arXiv). This paper considers three “stability” problems in topological property testing: namely, problems of the form “are objects almost-X close to X”, where X (here) is one of homorphisms, coverings of a cell complex, or 1-cocycles. The main result of the paper is that the three property testing problems are equivalent: namely, they admit testing proximity-oblivious testers (POTs) with similar rejection probability rates.
Testing Higher-order Clusterability on graphs, by Yifei Li, Donghua Yang, and Jianzhong Li (arXiv). The authors propose a new notion of graph clusterability, higher-order clusterability, meant to generalize the previously studied notions of clusterability; and proceed to provide testing algorithms for this notion.
Private Distribution Testing with Heterogeneous Constraints: Your Epsilon Might Not Be Mine, by Clément Canonne and Yucheng Sun (arXiv). This distribution testing paper (carry-over from the previous month) focuses on the extensively studied problem of closeness testing: given samples from two unknown distributions \(p\) and \(q\), decide whether \(p=q\) or if they are far. Now, add (differential) privacy: what if the two sets of samples, from \(p\) and \(q\) respectively, were sensitive data? And now, the focus of the paper… what if the two sets of samples were not equally sensitive? What are the trade-offs between the number of samples from \(p\) and the number from \(q\), then?