Fuzzy Selection of Local Search Sites in the Bees Algorithm
Authors: D.T Pham, Ahmed Haj Darwish
Abstract
The Bees Algorithm is a swarm-based algorithm that mimics the natural food foraging behaviour of honey bees. The algorithm essentially involves both random exploration of the solution space and more focused exploitation of promising local search sites. A basic version of the Bees Algorithm has been applied to function optimisation and a variety of other practical problems. This paper describes an enhanced version implementing fuzzy greedy selection of local search sites. The paper presents the results obtained for a number of benchmark problems to demonstrate the robustness and self-organising ability of the new Bees Algorithm.
| Attachment | Size |
|---|---|
| iproms.wmv | 5.95 MB |
Hello authors,
If I am not mistaken, the intention of this work is to improve the number of parameter settings for the Bees Algorithm. It looks promising.
I would like to know is there any drawback compare to the Basic Bees? It could be in terms of computing time, extra program codes, etc.?
Is there any hidden parameters need to be set in Fuzzy Greedy Selection system?
Do the natural bees exhibit behaviour similar to Fuzzy Greedy Selection System?
Thanks
ang
Hi
Thanks for your question
The selection process is done by zero order sugeno fuzzy system with weighted average, this type of fuzzy system is compact and it can be built by a combination of simple mathematical operations( sum , multiply and division). So it is not that complex or difficult to implement and it requires a little computing time more than the basic Bees Algorithm.
The natural behaviour of bees is a greedy behaviour, where bees always leave poor flower sites, and there is no hard threshold of the process selection.
Hello,
Thanks for your answer. Can you explain further what do you mean by 'the system is compact'?
Also, is there any parameter in the fuzzy system, if yes, how many parameters needed?
I also noted in your paper in section 3.2, the paragraphs need some attention. Refer the quote from your paper as below:
........... from the candidate
list.
This hierarchical abandonment is applied...
Regards
ang
Hi
There are two type of fuzzy inference systems, mamadani and sugeno, compact is here to show that sugeno type is compact compared with mamdani system , and it shows that it require less code and computation resources.
Candidate list is the list which stores all recent local search sites.
Hi
For the fuzzy system ,it need only rank value and fitness value for input and the number of workers for output
Dear Ahmed,
I would like to let you know that on page 4 of your paper (at section 3.2.), there are formatting errors. I hope you could rectify them soon. In the second paragraph of the section 3.2., You have forgotten the letter "G" for the name of Ghanbarzadeh and the third paragraph has no indentation for a new paragraph.
Thank you.
Regards
Ng
Sameh Otri
PhD Research Student
I understand that the fuzzy system needs these two values: ranking & fitness of each patch, then the system will tell you how many bees to recruit for that particular patch. Originally, in our Bees Algorithm, this number of recruitments is given a simple nsp will you give it a symbol (m) in the modified Fuzzy Bees Algorithm. Or it is not like that. Correct me if I am wrong.
The foraging behaviour of honey-bees is also a greedy because the honey-bees look for a good patch, utilise it then move to the next one but not to the best patch.
so Bees, in thier daily life, do not perform an optimisation technique. while in our case, as engineers, we are trying to find the optimum solution.
Sameh Otri
thanks for your prompt reply..
On section 3.1.2., it was stated:" A greedy algorithm leads to an optimal solution because a locally optimal choice leads to a globally optimal solution.
Do you have any prove of this statement?
It it not always true!!!
I cannot see any link between locally optimal choice and a greedy algorithm
You may choose a greedy local step which is optimal but in a global view it is not. This is way in research a greedy step leads to local optimal and mat trapped in it?
Hello
Local optimum solutions set consists of solutions which are local and may be global, so that means that global solutions are members in local solutions set and members in global solutions set.
If you like to get more information, read (Russell S. and Norvig P. Artificial Intelligence: A
Modern Approach. 2nd Edition. 2004: Prentice-Hall.)
In page 2, section 3, Recruit bees for selected patches (more bees for best patches) and evaluate their fitness.
Why you still keep the idea of recruiting more bees for best patches while your system work in fuzzy way and there are not such divisions.
I believe the pseudo code need to be revised to reflect the why the Fuzzy bees works?
Thx
Hi
there is no division in recruitment process, but if you go to page 4 you can find figure 5(the fuzzy system surface), it shows that recruitment increases for the best sites(best sites in terms of fitness and rank)
The proposed algorithm uses soft threshold to classify bad sites and good sites, instead of hard threshold in the basic Bees Algorithm to divide the visited sites in different groups.
Excellent and Thank you Ahmad for this original work!
Let me start with simple question to see if I understand your work correctly:
For simplicity lets call: Basic Bees Algorithm as (BBA) and Smart Bees Algorithm as (SBA)
The parameters for (BBA) are:
Number of scout bees (n); number of patches selected for the neighbourhood search (out of n visited patches) (m); number of top-rated (elite) patches among m selected patches (e); number of bees recruited for the best e patches (nep); number of bees recruited for the other (m-e) selected patches (nsp); the initial size of each patch (ngh) (a patch is a region in the search space that includes a visited patch and its neighbourhood) and the stopping criterion.
While the parameters for (SBA) are:
Number of scout bees (n) and maximum number of worker bees in each patch (m).
Because originally in (BBA) m patches are chosen out of n visited patches), while here in (BBA) : m worker bees are recruited for each scout bee n. am I right?
If yes. Calling the maximum number of worker bees in each patch as (m) in the (SBA) contradicts with the definition of m in the (BBA). Is it not more logic to call the maximum number of worker bees in each patch as (nep).
Sameh Otri
PhD Research Student