New Core?

[Sheen]

I noticed in the news section that a new core has been released. Do we as clients need to do anything, or will it update itself?

[Sheen]

After watching my states for a hour or twom I withdraw my last question, as it seems to be running much faster now.
:rolleyes:

[ashp]

I'm just curious as to what changed? Obviously a 20x increase in processing is a massive change. Just tweaked things, or moved to a whole new algorithm?

Also, I hate to ask an annoying question when I should be satisfied with the new core, but are we going to get http proxying anytime soon? At the moment I'm only running the code on a single machine, because of the forwarded ports, when I have an AMD 2500+ sitting idle, just waiting to be let loose. :)

[michaelgarvie]

I'm just curious as to what changed? Obviously a 20x increase in processing is a massive change. Just tweaked things, or moved to a whole new algorithm?
Basically the circuit mapping used to allow sequential circuits which meant they had to be simulated for enough time to be sure that its behaviour was characterised enough to give an accurate fitness measure. What's changed is that now all genotypes are mapped to combinational circuits. This means that if the gates are refreshed in the correct feed forward order (from primary inputs towards primary outputs), then they only need to be refreshed once. Hence simulation used to required 30 time steps but now only requires 1. There are reasons why allowing sequential circuits may help to make the fitness landscape smoother so helping evolution arrive at high fitness areas. However the speed advantage of forcing it to explore only combinational areas of the landscape may be larger. We shall see when the results start coming out!


Also, I hate to ask an annoying question when I should be satisfied with the new core, but are we going to get http proxying anytime soon? At the moment I'm only running the code on a single machine, because of the forwarded ports, when I have an AMD 2500+ sitting idle, just waiting to be let loose.
I know, you're not the only one who's requesting HTTP proxying. The situation right now is that I'm writing up my PhD thesis and consequently don't have any time on my hands. I would love to get this going and do know how to do it. If any experienced programmer feels like joining the project to contribute features such as this, or like including Effort Rate graphs in the client, they would be very welcome indeed. The payment would be great kudos within the community!