Fig 1.
Main step of original PO.
Fig 2.
(A) Population initialization, different shapes refer to different parties. Quadrilaterals, pentagons, and circles represent political parties P1, P2, and P3, respectively. Dotted frames C1, C2, and C3 represent different constituencies. (B) Members of various political parties conduct canvassing activities within their respective constituency. (C) Party leaders (solid) and constituency winners C1’, C2’, and C3’ are determined. (D) Update the positions of party members according to the constituency winners. (E) Update the positions of party members according to the party leaders. (F) The resultant positions are synthesized according to the position of the constituency winners and the party leaders. (G)(H) party switch, in constituency C3 is exchanged with
in constituency C1. (I) election phrase and reassign party leaders and constituency winners. (J) parliamentary affairs: each parliamentarian is updated if its fitness is improved after being attracted by a random parliamentarian. (K) final positions after one iteration.
Fig 3.
Example of interpolation.
Fig 4.
The QIPO process.
Fig 5.
Cases when QI and AQI deal with multimodal problems.
Fig 6.
The principle of refraction learning.
Fig 7.
The flow chart of CRLPO.
Table 1.
Main characteristics of 19 widely used benchmark functions.
Fig 8.
Convergence curve on 6 representative benchmark functions with 10 dimensions.
Fig 9.
Convergence curve on 6 representative benchmark functions with 30 dimensions.
Fig 10.
Convergence curve on 6 representative benchmark functions with 50 dimensions.
Fig 11.
Convergence graph of CRLPO on 4 representative benchmark functions with 10 dimensions.
Fig 12.
Convergence graph of CRLPO on 4 representative benchmark functions with 30 dimensions.
Fig 13.
Convergence graph of CRLPO on 4 representative benchmark functions with 50 dimensions.
Table 2.
Experimental results of 19 benchmark functions from Table 1 with 30 dimensions.
Table 3.
p-values of the Wilconxon rank-sum test at 0.05 significance level for CRLPO against other eight algorithms on 19 benchmark functions from Table 1 with 30 dimensions.
Fig 14.
Convergence graph of CRLPO and eight other algorithms on six representative benchmark functions with 30 dimensions.
Table 4.
Experimental results of 19 benchmark functions from Table 1 with 100 dimensions.
Table 5.
p-values of the Wilconxon rank-sum test at 0.05 significance level for CRLPO against other eight algorithms on 19 benchmark functions from Table 1 with 100 dimensions.
Fig 15.
Convergence graph of CRLPO as well as eight other algorithms on six representative benchmark functions with 100 dimensions.
Table 6.
Experimental results of 19 benchmark functions from Table 1 with 1000 dimensions.
Table 7.
p-values of the Wilconxon rank-sum test at 0.05 significance level for CRLPO against other eight algorithms on 19 benchmark functions from table 1 with 1000 dimensions.
Fig 16.
Convergence graph of CRLPO as well as eight other algorithms on six representative benchmark functions with 1000 dimensions.
Table 8.
Comparison of CRLPO and nine selected algorithms on 30 test functions with 30 dimensions from IEEE CEC 2014.
Table 9.
Experimental results of effectiveness evaluation of two components of CRLPO on 19 test functions from Table 1 with 100 dimensions.
Table 10.
Experimental results of CRLPO for 19 test functions with typical combinations of ξ and p(Dim = 30).
Table 11.
Experimental results of CRLPO for 19 test functions with typical combinations of ξ and p (Dim = 100).
Table 12.
Experimental results of CRLPO for 19 test functions with typical combinations of ξ and p (Dim = 1000).
Table 13.
Experimental results of RLPO with various interpolation strategies on 17 test functions from [51] (Dim = 10).
Table 14.
Experimental results of RLPO with various interpolation strategies on 17 test functions from [51] (Dim = 30).