SPSO2011
所属分类:数值算法/人工智能
开发工具:matlab
文件大小:14KB
下载次数:85
上传日期:2012-06-13 15:38:28
上 传 者:
hudieyutanglang
说明: 老外写的PSO算法,含基准测试函数,内容丰富
(PSO algorithm written by foreigners, including the benchmark function, rich in content)
文件列表:
SPSO2011\.DS_Store (6148, 2011-05-27)
SPSO2011\alea.m (99, 2011-05-23)
SPSO2011\alea_normal.m (532, 2011-05-23)
SPSO2011\alea_sphere.m (483, 2011-05-23)
SPSO2011\benchmark_func.m (28339, 2011-05-22)
SPSO2011\disp_fun.m (889, 2011-05-18)
SPSO2011\func.m (8414, 2011-05-24)
SPSO2011\get_fun_info.m (4423, 2011-05-19)
SPSO2011\SPSO2011.m (6091, 2011-05-27)
SPSO2011\test_SPSO2011.m (1214, 2011-05-27)
SPSO2011 (0, 2012-06-13)
{\rtf1\ansi\ansicpg1252\cocoartf1038\cocoasubrtf350
{\fonttbl\f0\fswiss\fcharset0 ArialMT;}
{\colortbl;\red255\green255\blue255;}
\paperw11900\paperh16840\margl1440\margr1440\vieww9000\viewh8400\viewkind0
\deftab720
\pard\tx566\tx1133\tx1700\tx2267\tx2834\tx3401\tx3968\tx4535\tx5102\tx5669\tx6236\tx6803\pardeftab720\ri380\ql\qnatural
\f0\fs20 \cf0 Developed by: Mahamed G.H. Omran (omran.m@gust.edu.kw) and Maurice Clerc \
27-May-2011\
\
To run SPSO2011:\
\
1) Copy all files into the Matlab directory\
2) Edit test_SPSO2011 to set the following parameters:\
a) Nr (Number of runs, default is 25)\
b) normalize (default is 0, no normalization)\
c) FE_max (number of function evaluations, default is 100,000).\
d) N (swarm size, default is 40)\
3) run test_SPSO2011, it will ask you for the function you want to optimize and the problem's dimension.\
\
\pard\pardeftab720\ri380\ql\qnatural
\cf0 Compatibility with the original C version\
\
1) RNG and numerical instability.\
\
We (Clerc and Omran) conducted several experiments on some benchmark functions. \
For several functions the results are very similar to the ones of \
the C version while for one function (CEC F6) the SR is really different:\
\
F6 results (C code):\
a) No normalization\
Avg. fitness = 5.69e+01 (2.06e+02)\
SR= 49.20 %\
\
b) Normalization\
Avg. fitness = ***4e+01 (1.57e+02)\
SR= 37.2 %\
\
F6 results (Matlab code)\
a) No normalization\
\
Avg. fitness = 6.05e+01(1.58e+02) \
SR = 0%\
\
b) Normalization\
\
Avg. fitness = 5.12e+01(1.37e+02) \
SR = 0%\
\
We suspect that there is a problem with the Matlab RNG and/or numerical \
instability (we implemented a simple RNG using C and Matlab and run our \
programs and still we got different results).\
\
2) Normalization\
\
It is recommended that you use this option (i.e. randomize = 1) when \
the search space in not a hypercube. \
If the search space is a hypercube, It is better not normalize \
(there is a small difference between the position without any normalisation and the de-normalised one.).\
\
3) Random permutation\
\
The random permutation of the numbering of the particles before \
each step is not included in the Matlab version(usually, it does not \
make a great difference in the C version).\
\
\
}
近期下载者:
相关文件:
收藏者: