Innovative Production Machines and Systems Conference

Well, why don't you try from here, or you can write down your question under one of the papers in this session.
Regards,
Ebubekir

Hi Dr.Dent

First of all, I believe it is real scientific honey not hype !!

For your nice comprehensive question, let me answer it in two sections. First comparison the Bees Algorithm with PSO, then GAs.

In PSO, as far as I know (I admit that I am not an expert in this area) global best, or in some versions, local bests are attraction other particles(in another word, they have influence to another particle movement). But, in the current version of the Bees Algorithm selected sites (local bests) have no influence to the movement of other bees, and only retrieved bees for those sites are influenced by the selected site.

Abut GAs, usually in GAs two parrents are chosen and with assist of some operators, children will be created, but in the Bees Algorithm only one sites will be selected and search will take place around selected sites. Although, GAs benefits from mutation that operates on one chromosome, but it is more like random search in the Bees Algorithm than neighbourhood search.

I am sure other bees and maybe wasps will help me to complete the answer.

Regards,

Afshin

Hi Afshin,

Many thanks for your reply, but I am still undecided about the honey/hype issue!

Here is why. I cannot see how having uncommunicative local bees will make the Bees Algorithm better than PSO.

Also (and I might have misread you), I cannot imagine random site searching by bees to be superior to GA mutation.

Regards,

Olivier

Where is the new nectar you have been promising us?

Dear Dr Dent,

Thanks for your questions again.

For the first part, one of the drawbacks in standard PSO is local optimums. It is more likely that PSO stuck in local optimum in some cases (i.e. problems have loads of local optima and especially if the global optimum is located in the corner of the search space). In the Bees Algorithms, as the selected sites do not have influences to each other so if some of them stuck in local optimums, they do not attract others to those local optimums. In the new version of the Bees Algorithm, selected sites which stuck in the local optimums, will be realised which I would like to refer you to the paper for more details (abandon sites without new information). Another issue is the NFL (No Free Lunch) theory, as we are claiming the Bees Algorithm is giving better results than PSO in many cases not all cases.

For the second part, what I meant, was random search part in the Bees Algorithm is almost similar to the mutation in GAs and both will help the algorithm escape from the local optimums.

Hope this answer helps you to decide!!!

Please do not hesitate to ask more crucial questions and hereby I would like to ask co-chair to encourage other bees to participate in this discussion.

Regards,

Afshin

Dear Dr Afshin,

Many thanks for your helpful reply!

The Bay Bees have decidedly created something very good here!

Congratulations!

Regards,

Olivier

Dear Bay Bees,

Why have you not responded to Afshin's cry for help?

Are you still beesily collecting nectar or have you all gone on vacation?

I know that Olivier Dent seems happy with Afshin's reply for now, but do not be complacent as he can strike again at any time.

Prepare yourselves to defend your nest against future attacks by Dr Dent and by wasps, ants, birds and other genetically different species !

Yours truly,

DTP.

Have no fear!

I am a friend of the Bay Bees', not a foe!

Like you, I am a peace-loving and hardworking creature.

I am not interested in attacking you nor anyone else.

My questions to you were sparked by a true sense of scientific curiosity: I simply was intrigued by your Bees Algorithm and wanted to satisfy myself that it was as good as you had claimed it to be.

You are quite right. I am happy with Dr Afshin's reply - for now.

When I have more questions for him or for your other Bees (do they really exist? I have heard from nobody apart from Dr Afshin), you can be sure that I shall get back to you!

Hi

We treat you as our friend not enemy, wasp or …. !!! Actually we are so happy from your questions as they will clarify the Bees Algorithm more and help others to understand it better as well.

Please come up with more questions and hereby I would like to invite others to write theirs questions, comments about the Bees Algorithm and the Cardiff Bay Bees will be so happy to answer the questions and I am sure it will help the team to improve the algorithm even more.

Regards,

Afshin

Hi

You are so kind. Thanks.

As the Bay Bees are my friends, I have tried hard to come up with a few straightforward questions for you all. So here goes.

1. How easy is it to choose the values of the Bees Algorithm parameters?

2. How sensitive is the performance of the Bees Algorithm to those values?

3. Do you have guidelines to help users to select an acceptably good set of values?

4. Do those values differ markedly for different problems?

Regards

MRO Dent

Hi

Thanks again for your nice questions, actually the answers to these questions, will give a very good general knowledge about the Bees Algorithm.

I should admit that in the first version of the Bees Algorithm, the performance of the Algorithm was sensitive to the parameters of the algorithm, but in the current version because of the "Shrinking" and "Abandon sites without new information" (details can be seen in the paper titled: Multi Objective Optimisation using the Bees Algorithm and Preliminary Design using the Bees Algorithm in Lamdamap 2007 conference) the algorithm is not that sensitive as the neighbourhood size will be adapted for each site regarding the progress of the bees searching in that site.

Setting the parameters of the algorithm is not difficult and users can guess what are the best values to start especially if they are familiar with evolutionary algorithms. As I mentioned in the first part, if the user do not choose the best set of the parameters, the algorithm in most cases can still reach to the optimum, of course with more sampling (evaluation).

I would like to say that a paper will be published soon about the statistical analysis and effects of the parameters of the Bees Algorithm and that paper will give a very good guideline about what is the best range of the parameters to reach to good results and also, what is the range of the parameters which should be avoided as they will end to poor results. So, bear with us and I will promise you will see a nice analysis (I would like to thank, visitor bee for this work).

About the last question, regarding my experiences of working and solving different problems in different categories, the parameters are depended to number of variables in the problems and less depend on the category of the problems. Of course some exceptions exist as in all algorithms and engineering domain.

I hope other bees share their experiences to have a good conclusion in the end.

By the way, where are the Bay Bees Gone!!!!????

Afshin

Hello Afshin,
Many thanks for your counter attack.
I hope it will silence our "friend" for good!
Best wishes.
DTP.
PS: You are right! Where are the Bay Bees gone??? Where are the Bay Bees gone??? Hopefully, not too far away!

Thanks Afshin for replying and defending the Bees.

You are the General Bee and you are doing very well.

Please call for help from other Bees when you need.

So far I can see you are able to manage.

For combinatorial problems in general the number of parameters has dropped from 6 to 3 which are the total number of bees, the best selected bees and the number of recruited bees only. The performance of the Algorithm is not that sensitive
As well as, setting the parameters of the algorithm is not difficult with some trial and error.
At the moment, there is further study on providing users with guidelines to help them select an acceptably good set of values.

As the algorithm not so sensitive to the values of those parameters they do not differ markedly for different problems

Sameh Otri
PhD Research Student

Hi,

Thanks a lot for your answer.

By the way, why do you (and other researchers) want to copy bees, ants, etc?

Are we humans no good?

Regards,

MRO Dent

Hi Dr Dent,
Thanks for your nice comments and questions on our bees algorithm.
Regarding your latest question, its not about being good or not. I think, it is mostly about being simple or not!
We humans have practicle solutions to deal with any kind of problems. But this solutions comes from quite complex ways which is not easy to model.
For instance, "logic" is in one hand and brain is in onother. if you approach from "logical" point of view, you will (or may) end up in a endless progress like Aristotelian logic or modern logicians like Wittgenstein and others. However, if you go to brain side which is more pragmatic, you may have some models, like artificial neural networks, but again you may end up far from the real brain. One of the obvious reason is that the black box structure where there is not any explenation to your propositions why you get that result and so on.
On the other hand, there are other creatures who can deal with similar kind (ofcourse not that complex) of problems using some very simple reasoning or direct/inderect ways of communication techniques. These reasonings and/or communication techniques relatively easy to model and sometimes more effective than human-related techniques. Especially in optimisation field, literally they have more experience than us. Limited reasoning push them to produce simple relations and rules to deal with the complex problems for millions of years. This simple features are clearly more suitable for modelling...
Well, sorry for all this philosophical explanations but if you begin to talk about human being therefore you have to turn this direction as well.

Best Regards,
Ebubekir Koc

Hi,

I am new to this area and am fascinated by all the discussion going on here.

I support Mr Dent, though.

What can bees teach us that we don't already know?

The Bees Algorithm seems to employ a classical idea in optimisation: combining exploration (scout bees) and explotation (recruited bees).

What's really new?

Best regards,

A Newby

Hi,
Thanks for your nice question...
As you said the bees algorithm employs some of the classical ideas for optimisation.
Well, the answer for your question is also the answer for how to do it? and how to combine them? The new thing here is a new method that can find a reasonable balance in these two concepts (exploration and explotation). As you know, these are some of the essencials for optimisation. Any optimisation algorithm must use these methods alone or together in anyway. The question is how? So we have introduced a new method for this question inspiring from the bees foraging behaviours.
Best Regards,
Ebubekir

Hello Ebubekir,
Many thanks for helping to defend the hive!
See you at the Bees Council!
Best wishes.
DTP.

Hi,

Thanks for your nice reply. I have some more questions.

What is meant by "reasonable balance"? How does the BA achieve "balance"? How do we know it is "reasonable"?

Would "balance" not be dependent on the stage of the optimisation process?

Does the BA automatically find the optimum degrees of exploration and exploitation for the particular stage of optmisation?

Best Regards,

A Newby

Dear Dr Newby,
Thank you for your challenging questions.
Our Bay Bees are busy collecting nectar and making honey.
They will attend to your questions shortly.
Best wishes.
D Pham.

In addition of what Ebu Baker was trying to say:

Bees, ants and any other swarm insects are simple models and when it comes to real implementation as robotics designers it is much easier to model, design and test these simple creations rather than a complex system like human.
As well as, it is much cheaper and practical.

For your reference you can read:
(Swarm Intelligence from nature to artificial systems) by Bonabeau and others
[Q335.B6]

Sameh Otri
PhD Research Student