Feature Interaction in Terms of Prediction Performance

There has been considerable development in machine learning in recent years with some remarkable successes. Although there are many high-performance methods, the interpretation of learning models remains challenging. Understanding the underlying theory behind the specific prediction of various models is difficult. Various studies have attempted to explain the working principle behind learning models using techniques like feature importance, partial dependency, feature interaction, and the Shapley value. This study introduces a new feature interaction measure. While recent studies have measured feature interaction using partial dependency, this study redefines feature interaction in terms of prediction performance. The proposed measure is easy to interpret, faster than partial dependency-based measures, and useful to explain feature interaction, which affects prediction performance in both regression and classification models.

Exploring the imprint of social media networks on neighborhood community through the lens of gentrification

Gentrification, the rise of affluent socioeconomic populations in economically depressed urban neighborhoods, has been accused of disrupting community in these neighborhoods. Social media networks meanwhile have been recognized not only to create new communities in neighborhoods, but are also associated with gentrification. What relation then does gentrification and social media networks have to urban communities? To explore this question, this study uses social media networks found on Twitter to identify communities in Washington, DC. With space-time analysis of 821,095 geo-tagged tweets generated by 77,528 users captured from 15 October 2015 to 18 July 2016, we create a location-based interaction measure of tweets which overlays the social networks of the comprising users based on their followers and followees. We identify gentrifying neighborhoods with the 2000 Census and the 2010–2014 American Community Survey at the block group level. We then compare the density of location-based interactions between gentrifying and nongentrifying neighborhoods. We find that gentrification is significantly related to these location-based interactions. This suggests that gentrification indeed is associated with some communities in neighborhoods, though questions remain as to who has access. Making novel use of big data, these results demonstrate the important role built environment has on social connections forged “online.”

A consistent measure for lattice Yang–Mills

The construction of a consistent measure for Yang–Mills is a precondition for an accurate formulation of nonperturbative approaches to QCD, both analytical and numerical. Using projective limits as subsets of Cartesian products of homomorphisms from a lattice to the structure group, a consistent interaction measure and an infinite-dimensional calculus have been constructed for a theory of non-Abelian generalized connections on a hypercubic lattice. Here, after reviewing and clarifying past work, new results are obtained for the mass gap when the structure group is compact.