VENEZIA_SpectralClustering_Graph_Comp
Spectral Clustering of Urban Networks
Based on the map of Venice this is the plot of the network spectrum in three dimensions divided into six clusters.
Spectral clustering is particularly useful for analyzing data where where topological links outweigh convex boundaries, that is, when connections a more significant than simple proximity. This makes it an interesting method for the analysis of urban networks where cul-de-sacs, dead ends, physical barriers, or topographic distance can separate two points that are otherwise quite proximate.
Cluster centrality is shown on this plot by distance from the center point of each cluster the distribution of cluster centrality in the plot at left (with box plot around quartiles). Network links are not represented in this plot but comparing to the map view, one can quite easily intuit the relationship of nodes to their spatial location.
The final network comprises over 11,300 nodes and 12,500 edges.
We will be conducting a workshop at the Singapore pavilion of the Venice Biennale engaging this topic and will hopefully find modifications to this network and to the methods.
VENEZIA_SpectralClustering_Graph_Comp
Spectral Clustering of Urban Networks
Based on the map of Venice this is the plot of the network spectrum in three dimensions divided into six clusters.
Spectral clustering is particularly useful for analyzing data where where topological links outweigh convex boundaries, that is, when connections a more significant than simple proximity. This makes it an interesting method for the analysis of urban networks where cul-de-sacs, dead ends, physical barriers, or topographic distance can separate two points that are otherwise quite proximate.
Cluster centrality is shown on this plot by distance from the center point of each cluster the distribution of cluster centrality in the plot at left (with box plot around quartiles). Network links are not represented in this plot but comparing to the map view, one can quite easily intuit the relationship of nodes to their spatial location.
The final network comprises over 11,300 nodes and 12,500 edges.
We will be conducting a workshop at the Singapore pavilion of the Venice Biennale engaging this topic and will hopefully find modifications to this network and to the methods.