AU C heatmaps for games played on Erd˝ os-Rényi random graphs. Each row corresponds to a different connection probability p ∈ {0.04, 0.18}, while the columns refer to the rule parameter used by the agents β ∈ {1, 10}.

Contexts in source publication

Context 1

... plotted as a heatmap on an S-T plot ( fig. 3, fig. 4 and fig. 5). Here S and T refer to the parameters which specify a game as in eq. (1). In other words, each point in the plot corresponds to a different game with the color roughly characterizing the effectiveness of φ ij as a predictor for the existence of the edge (i, j) in the ...

Context 2

... about the topology of the graph. If this analysis is correct, we should expect to see that increasing the rule parameter β allows the agents to differentiate smaller payoff differences, and the quadrants of these heatmaps should tend to become more homogeneous. This is exactly what we see, particularly when edge density of the graphs is low (fig. 3, fig. 4 and fig. ...

Context 3

... brings us to the final factor involved in determining the efficacy of the classification: the edge density of the graph. When we compare, for example, the Barabási-Albert graphs with m = 1 ( fig. 3(a-b)) to the Erd˝ os-Rényi graphs with p = 0.04 ( fig. 4(a-b)), we see almost identical structure in the AU C heatmaps. On average the Erd˝ os-Rényi graphs have the same edge density to their Barabási-Albert counterparts, i.e. they have about the same number of edges. The cycle graph ( fig. 5(a- b)) has the same edge density to the Barabási-Albert graph (m = 1), and again we see almost identical ...

Context 4

... the same number of edges. The cycle graph ( fig. 5(a- b)) has the same edge density to the Barabási-Albert graph (m = 1), and again we see almost identical heatmaps. By increasing the edge parameters for the Barabási-Albert and Erd˝ os-Rényi graphs, m and p respectively, we can see a similar comparison for yet denser graphs ( fig. 3(c-d) and fig. 4(c-d)). Again, the Barabási-Albert (m = 5) ...

Context 5

... os-Rényi (p = 0.18) graphs have similar edge density. What's more, the same qualitative change is seen for a single type of topology when the edge density changes. For example, comparing fig. 3(a) to 3(c) or fig. 4(a) to 4(c), we see fewer harmony and prisoner's dilemma games admit a faithful inference. The same phenomenon is observed in comparing cycles, lattices and wheels ( fig. 5), which further supports the claim. The graphs have remarkably distinct topologies even with only 50-nodes. For example, it is difficult to find any similarity ...

Context 6

... plotted as a heatmap on an S-T plot ( fig. 3, fig. 4 and fig. 5). Here S and T refer to the parameters which specify a game as in eq. (1). In other words, each point in the plot corresponds to a different game with the color roughly characterizing the effectiveness of φ ij as a predictor for the existence of the edge (i, j) in the ...

Context 7

... about the topology of the graph. If this analysis is correct, we should expect to see that increasing the rule parameter β allows the agents to differentiate smaller payoff differences, and the quadrants of these heatmaps should tend to become more homogeneous. This is exactly what we see, particularly when edge density of the graphs is low (fig. 3, fig. 4 and fig. ...

Context 8

... brings us to the final factor involved in determining the efficacy of the classification: the edge density of the graph. When we compare, for example, the Barabási-Albert graphs with m = 1 ( fig. 3(a-b)) to the Erd˝ os-Rényi graphs with p = 0.04 ( fig. 4(a-b)), we see almost identical structure in the AU C heatmaps. On average the Erd˝ os-Rényi graphs have the same edge density to their Barabási-Albert counterparts, i.e. they have about the same number of edges. The cycle graph ( fig. 5(a- b)) has the same edge density to the Barabási-Albert graph (m = 1), and again we see almost identical ...

Context 9

... the same number of edges. The cycle graph ( fig. 5(a- b)) has the same edge density to the Barabási-Albert graph (m = 1), and again we see almost identical heatmaps. By increasing the edge parameters for the Barabási-Albert and Erd˝ os-Rényi graphs, m and p respectively, we can see a similar comparison for yet denser graphs ( fig. 3(c-d) and fig. 4(c-d)). Again, the Barabási-Albert (m = 5) ...

Context 10

... os-Rényi (p = 0.18) graphs have similar edge density. What's more, the same qualitative change is seen for a single type of topology when the edge density changes. For example, comparing fig. 3(a) to 3(c) or fig. 4(a) to 4(c), we see fewer harmony and prisoner's dilemma games admit a faithful inference. The same phenomenon is observed in comparing cycles, lattices and wheels ( fig. 5), which further supports the claim. The graphs have remarkably distinct topologies even with only 50-nodes. For example, it is difficult to find any similarity ...