Figure 1.
Probabilistic Beam Search Algorithm.
Figure 2.
Reduction Algorithm.
Figure 3.
Perturbation and Improvement Algorithm.
Figure 4.
Multilevel Probabilistic Beam Search Algorithm.
Table 1.
Experimental results for RandomSet instances.
Table 2.
MPBS and IBS comparison for RandomSet instances.
Figure 5.
Solution quality over time for different DNA datasets in LargeRealSet.
Figure shows evolution along execution time of statistical values (average and median) for relative percentage difference (RPD) from best-known solutions of solutions obtained by MPBS algorithm. Average solution obtained by IBS is also shown. Note that the value of solution is considered to be before an algorithm has provided its first solution, hence beginnings of average curves indicate the moment at which at least one solution has been obtained for all executions and instances in the dataset.
Table 3.
Experimental results for DNA LargeRealSet instances.
Table 4.
MPBS and IBS comparison for DNA LargeRealSet instances.
Figure 6.
Solution quality over time for different protein datasets in LargeRealSet.
Figure shows evolution along execution time of statistical values (average and median) for relative percentage difference (RPD) from best-known solutions of solutions obtained by MPBS algorithm. Average solution obtained by IBS is also shown. Note that the value of solution is considered to be before an algorithm has provided its first solution, hence beginnings of average curves indicate the moment at which at least one solution has been obtained for all executions and instances in the dataset.
Table 5.
Experimental results for protein LargeRealSet instances.
Table 6.
MPBS and IBS comparison for protein LargeRealSet instances.
Figure 7.
Sensitivity analysis for (top) and
parameters (bottom).
Figure shows average relative percentage difference (RPD) from best-known solutions along with corresponding confidence intervals at the 95% confidence level obtained by running MPBS algorithm with different settings for (beam width) and
(number of dominators) parameters for the initial construction of solution phase. Experiments were carried out on RandomSet instances with different alphabet sizes (
).
Figure 8.
Sensitivity analysis for (top) and
parameters (bottom).
Figure shows average relative percentage difference (RPD) from best-known solutions along with corresponding confidence intervals at the 95% confidence level obtained by running MPBS algorithm with different settings for (beam width) and
(number of dominators) parameters for the reduction of solution phase. Experiments were carried out on RandomSet instances with different alphabet sizes (
).
Figure 9.
Sensitivity analysis for (top) and
parameters (bottom).
Figure shows average relative percentage difference (RPD) from best-known solutions along with corresponding confidence intervals at the 95% confidence level obtained by running MPBS algorithm with different settings for (beam width) and
(number of dominators) parameters for the perturbation of solution phase. Experiments were carried out on RandomSet instances with different alphabet sizes (
).
Figure 10.
Sensitivity analysis for Probabilistic Beam Search algorithm considering different beam widths (top) and executions with different number of iterations (bottom).
Figure shows average relative percentage difference (RPD) from best-known solutions along with corresponding confidence intervals at the 95% confidence level obtained by running a single iteration of PBS algorithm with different beam widths and for best solution found after executing different number of iterations of PBS algorithm (with a fixed beam width of 100). Experiments were carried out on RandomSet instances with different alphabet sizes ().