Loading ...
Sorry, an error occurred while loading the content.

Messages List

6861

Blackboards

In behavior trees for computer games there is a "blackboard" where actions are recorded and used to make further decisions. I suppose something like that
sean_c4s
Aug 17
#6861
 
6860

Re: DDFA saving the champion

Thank you Greg we figured out how to generate new features on a new data set and now we are experimenting with this approach. We are starting to understand
garnier.juan@gmail.com
Aug 15
#6860
 
6859

Re: DDFA saving the champion

Hello Greg, Thank you for the response. I have a few questions, an input would be appreciated. 1) Just so I understand, the networks in the archive folder the
garnier.juan@gmail.com
Aug 13
#6859
 
6858

Re: DDFA saving the champion

Hi Juan, Do you still need assistance with getting what you need from the DDFA code? There isn't really a concept of a champion with DDFA, since each
thebigo1999
Aug 12
#6858
 
6857

Re: DDFA saving the champion

Thank you Ken for your response I was able to use the network save and a bpw to load the network.
garnier.juan@gmail.com
Aug 12
#6857
 
6855

Re: DDFA saving the champion

Hello Ken, Let me try to explain as there are lots of things in this code that are very new to me so please excuse my amateurism here lol. In regards to the
garnier.juan@gmail.com
Aug 11
#6855
 
6854

Re: DDFA saving the champion

Just to clarify, what do you mean by "current champion" with respect to DDFA? DDFA collects a whole bunch of small networks based on novelty for later
kenstanley01
Aug 11
#6854
 
6853

DDFA saving the champion

Hello everyone wanted to know if anyone has experience with the DDFA code and would know how to save the best current champion to a folder to load later. I am
garnier.juan@gmail.com
Aug 10
#6853
 
6851

Re: Minimal Criterion Coevolution Experiments?

Hello Ken, My colleague and I are working with the DDFA code and the MNIST data experiment. As you know in that code the behaviors that Novelty Search
garnier.juan@gmail.com
Aug 1
#6851
 
6850

Re: Minimal Criterion Coevolution Experiments?

Hi Juan, I'm interested to hear you're thinking of combining HyperNEAT with MCC. I'm curious what you mean when you mention increasing "accuracy?" What does
kenstanley01
Jul 31
#6850
 
6849

Hyper Networks

Interesting blog post and paper with HyperNEAT references. From September 2016, but popular on hacker news today:
Colin Green
Jul 30
#6849
 
6848

Re: Minimal Criterion Coevolution Experiments?

Thank you very much Drew for you code and research paper I will look through it. I will keep you posted on the progress. Thank you again.
garnier.juan@gmail.com
Jul 25
#6848
 
6847

Re: Minimal Criterion Coevolution Experiments?

Thank you very much Jonathan for the response and great insight. My team and I will experiment with MCC and Hyperneat to see if this approach could potentially
garnier.juan@gmail.com
Jul 25
#6847
 
6846

Re: Minimal Criterion Coevolution Experiments?

Juan, Not sure if this helps at all, but my thesis work had a C++ implementation of competitive coevolution that used Minimal Criterion in some of the novelty
Drew Kirkpatrick
Jul 23
#6846
This message has attachments
  • PDF
    1.2 MB
    Thesis- Master Document.pdf
  • 710 KB
    drewK_thesis.tar.bz2
6845

Re: Minimal Criterion Coevolution Experiments?

Hi Juan, HyperNEAT excels at exploiting geometric properties of a given domain. In the MCC experiments as well as most of the original novelty search
Jonathan Brant
Jul 23
#6845
 
View First Topic Go to View Last Topic