Figure 4

Speed and storage improvement by distributed computing and rotational algorithm and parameter impact on algorithm performance. (A) The speed-up ratio of rotational tensor-vector product strategy over normal one. The left figure is about convergent time run with normal strategy. Its x-axis represents convergent time at unit of millisecond. The middle figure is about convergent time run with rotational strategy. Its x-axis represents convergent time at unit of millisecond. The right figure is about speed-up ratio run of rotational strategy. Its x-axis represents the corresponding speed-up ratio of rotational strategy over normal one. Their y-axis represents 15 different network pairs. (B) The speed-up ratio of distributed rotational tensor-vector product method over stand-alone one. The left figure is about convergent time run with distributed rotational strategy. Its y-axis represents convergent time at unit of millisecond. The middle figure is about convergent time run with stand-alone rotational strategy. Its y-axis represents convergent time at unit of millisecond. The right figure is about speed-up ratio of distributed rotational strategy. Its y-axis represents the corresponding speed-up ratio of distributed rotational strategy over stand-alone one. Their x-axis represents 6 different network pairs. (C) Relationship between α and alignment speed and accuracy. The x-axis is α value and in logarithmic coordinates. The first figure is about the alignment score after discretization. Its y-axis represents relative alignment score after discretization normalized by the corresponding values at α = 0.0001. The second figure is about the convergence time. Its y-axis represents relative convergence time normalized by the corresponding values at α = 0.0001. The third figure is about the edge correctness. Its y-axis represents relative edge correctness normalized by the corresponding values at α = 0.0001. The fourth figure is about the vertex correctness. Its y-axis represents relative vertex correctness normalized by the corresponding values at α = 0.0001. (D) Relationship between λ and alignment speed and accuracy. The x-axis is λ value. The first figure is about the alignment score after discretization. Its y-axis represents relative alignment score after discretization normalized by the corresponding values at λ = 0.9. The second figure is about the convergence time. Its y-axis represents relative convergence time normalized by the corresponding values at λ = 0.9. The third figure is about the edge correctness. Its y-axis represents relative edge correctness normalized by the corresponding values at λ = 0.9. The fourth figure is about the vertex correctness. Its y-axis represents relative vertex correctness normalized by the corresponding values at λ = 0.9. The letters in legend of C and D represents 15 different network pairs. Please see Supplementary Data 2 for the network pairs used in these figures.