Skip to main content

Advertisement

Fig. 6 | BMC Systems Biology

Fig. 6

From: Designing synthetic networks in silico: a generalised evolutionary algorithm approach

Fig. 6

Assessing the influence of different algorithm settings on performance. a-c Optimal simulations (red lines: after 2500 generations; green lines: after 40000 generations) are compared to the input time-series data (blue lines) for the three objective functions used. (A) Optimisation where network size and connections are fixed. b Optimisation where connections between nodes is left free. c Optimisation where network size and connections can be chosen freely. d Algorithm settings were randomly selected 1933 times and the convergence between initial and final scores was recorded (see Additional file 1). The convergence score for each property was then averaged across all EA simulations and normalised to property 11. Property 1 = ‘Probability of gene addition’; 2 = ‘Network mutation rate’; 3 = ‘Probability of moving a connection’; 4 = ‘Probability of deleting a connection’; 5 = ‘Probability of gene mutation’; 6 = ‘Maximum number of offspring’; 7 = ‘Optimisation objective’; 8 = ‘Parameter mutation method’; 9 = ‘Probability of adding a connection’; 10 = ‘Parameter mutation rate’; 11 = ‘Selection method’. e The best and worst performing 25 EA runs were analysed to determine which selection method was used in the EA run

Back to article page