Thank you sir for this presentation. I have one question, please Is DE/current to best is good for exploration or for exploitation or it offers a balance between the two?
Great question. Like most DE implementations it tries to strike a balance between the two. The difference vector between "best and parent" tries to explore more where the parent is weak (relative to best particle) and smaller, i.e. more exploitation where the parent the best particle are closer to one another.
Good revision video for my exam tomorrow, thank you🔥🙏🏾 Are you by any chance related to Prof Theo van Niekerk, who lectures at NMU? You sound slightly like him, and you give good explanations also! He taught me Mechatronic control systems
Thanks for the vid! Doing this for a school assignment right now and your explanation was very helpful.
I am really glad. Good luck and keep on learning.
Is it possible to have the presentation file of this video?
excellent explanation
Thank you sir for this presentation.
I have one question, please
Is DE/current to best is good for exploration or for exploitation or it offers a balance between the two?
Great question. Like most DE implementations it tries to strike a balance between the two. The difference vector between "best and parent" tries to explore more where the parent is weak (relative to best particle) and smaller, i.e. more exploitation where the parent the best particle are closer to one another.
@@drjacovanniekerk Could you please make another video explaining this point, please?
@@drjacovanniekerk So, can we say it is more exploitative?
@@drjacovanniekerk What if the best is far away to the global mimimum for example? For DE/current to best
@@drjacovanniekerk If they the parent and the best are far away to each other. Does it mean that DE/current to best is explorative?
Hello, can you share book name?
Sorry, I missed this message. The book is: Computational Intelligence: An Introduction, Second Edition, Andries P. Engelbrecht
9:20 should the difference vector be the magnitude (i.e. always positive)?
No, you are adding vectors, not scalars thus you need to preserve the directional information to navigate through search space.
Good revision video for my exam tomorrow, thank you🔥🙏🏾
Are you by any chance related to Prof Theo van Niekerk, who lectures at NMU? You sound slightly like him, and you give good explanations also! He taught me Mechatronic control systems
Probably... but I really don't know how as I do not know him. :-) We're all technically related. Thank you for the compliment.
Is the target vector the same for each parent or do we use a specific trail vector for each parent?
A new target vector is chosen for each parent.