RE: Is there objective Truth?
October 28, 2016 at 4:12 am
(This post was last modified: October 28, 2016 at 4:16 am by I_am_not_mafia.)
(October 26, 2016 at 4:59 pm)Soldat Du Christ Wrote: Gene duplication isn't new information being added, it's duplication.
You're right, but the neutral gene theory explains this as opening up new areas of search space. Neutral duplications that do not reduce the fitness of an organism allow for mutations in future generations to be applied to the duplicated part. Therefore duplication allows for new information to be added in the future. I use this myself in my evolutionary algorithms to let more complex solutions develop over time.
(October 26, 2016 at 4:59 pm)Soldat Du Christ Wrote: "Mutations leading to new information (e.g., way of functioning, physical trait). Sometimes, it can have negative effects, sometimes positive, and sometimes a combination of both, depending on the environment type in which the mutation (or set of mutations) develop(s)."
Convergence loses information over time. This will happen even if you do not have a selection pressure. Mutation increases information over time.
With genetic algorithms you need to carefully set your mutation rate. Set too high and you will find a solution more quickly, but at the cost of finding a near optimal solution. Set it too low and you will spend longer trying to find a good solution.
This is why I personally developed an evolutionary algorithm that does away with convergence and only uses the mutation operator and decreases the mutation rate over time.
This is all theory put into practice when writing genetic algorithms. This is theory developed from finding out what actually works in practice. If it was wrong, our algorithms wouldn't work. I can dig out the sources if you want.
