Skip to main content

Advertisement

Table 3 Optimising the model to experimental saccadic velocity profiles: fitness values on individual objectives

From: Optimisation of an exemplar oculomotor model using multi-objective genetic algorithms executed on a GPU-CPU combination

Selection method 5 10 20
I 16.6687 (0.0229) 16.5143 (0.0190) 13.9354 (0.0125)
II 3.1672 (0.0285) 94.7590 (0.0072) 138.2400 (0.0044)
III 40.7405 (0.1000) 8.7522 (0.0557) 36.6450 (0.1692)
IV 19.1386 (0.0202) 17.3887 (0.0106) 13.1121 (0.0013)
  1. Results are shown for each method used to select the final solution. Method I selects the solution that minimises the Euclidean
  2. distance between the Pareto front and the objective space origin; method II selects the best fit to a 5 deg saccade (objective 1);
  3. method III selects the best fit to a 10 deg saccade (objective 2); and method IV selects the best fit to a 20 deg saccade
  4. (objective 3). Mean fitness values and coefficients of variation (shown in brackets) were calculated from 16 runs of NSGA-II
  5. for a population size of 8000. Fitness values were normalised by the number of points in each velocity profile