openmolecules.org

 
Home » DataWarrior » Bug Reports » evolutionary algorithm
evolutionary algorithm [message #480] Thu, 21 February 2019 22:59 Go to next message
pc419714@ohio.edu is currently offline  pc419714@ohio.edu
Messages: 20
Registered: February 2019
Junior Member
Thomas,

Could you describe how the evolutionary algorithm was written? I have some scaffolds that it evolved, however I am wondering why the fitness sometimes appears to decrease before increasing in a stepwise fashion. If I had it set up to generate 4096 compounds per generation, unlimited generations, with 32 compounds surviving each generation, and generate compounds similar to my scaffold without anything in the structure selected, how does it determine which compounds to start the next generation with? It just appears to increase and then decrease sometimes.... I thought it would just keep increasing in fitness every time.... Does it just improve upon the lower fitness molecules from the previous generation? How is fitness defined? I read the documentation. I just wondered if you used a formula or anything to make the algorithm.

Thank you

Patrick Chirdon
Re: evolutionary algorithm [message #489 is a reply to message #480] Thu, 07 March 2019 17:52 Go to previous message
thomas is currently offline  thomas
Messages: 716
Registered: June 2014
Senior Member
Hi Patrick,

during the creation of the new generation structures DataWarrior does not look at any fitness criteria. It just creates compounds with small changes from the previous generation, which have not been generated in any of the previous generations. Thus, it may be that the best compound of the generation has a worse fitness than the best compound of the previous generation.

Thomas
Previous Topic: Issues with multiple columns of smiles in 5.0.0
Next Topic: Mapping tool defective
Goto Forum:
  


Current Time: Thu Dec 26 14:31:18 CET 2024

Total time taken to generate the page: 0.02676 seconds