What you see.

[michaelgarvie]

When running the DHE client from a script you can see:
A. In the console window the history of the best individual of the generations with its genotype and fitness scores at the main task <f>, off-line BIST <p0>, on-line BIST <p1> and parsimony <p2>.

B. In the GUI.
1. A graph with your neighbours and the fitness scores of their best individuals.
2. A description of the current run.
3. The current generation and its current best individual.
4. A circuit diagram of the best circuit.
All this described in the FAQ: http://www.cogs.susx.ac.uk/users/mmg20/dhe.htm#gui

In the Statistics page you can see:
1. How many circuits you've evaluated.
2. How many days you've been submiting packets for.
3. How much computational effort all your circuits required to evaluate.
4. Your average circuits per day.
5. How many times an individual from your island has been the best of all islands.
6. All this info per team.
7. The relevant info for today and yesterday.
8. A fitness graphs showing how f,p0,p1 and p2 have been changing since the run started.
9. A graph showing the topology of islands and how fit their best individuals are.
10. A table with all results this project has produced which are better than human design, together with their circuit diagrams.

What else would you like?

[spaetz]

I would like to
a) be able to *remove* views from the GUI, i.e. be able to switch off the circuit diagram in order to save screen space. Now, all I can do is pull the thumb thingy as small as possible. Doesn't Swing provide a grippy that goes to minimum size when you click on it? Ahh, and if the view of my choice would be persistent across sessions that would be wonderful (but not absolutely necessary).

b) a graph which shows the fitness etc of my individuals over time (either in the GUI or the statistics website) (probably have the globally best individual drawn in there as well for a comparison on how good I am).

c) The ability to limit the timeframe of the statistical graphic which shows the numbers over time (on your website). (show last 10 days, 30 days, all) or something like that.

[michaelgarvie]

Changes can be seen on the hall of fame circuits:
http://www.cogs.susx.ac.uk/users/mmg20/stats.php#hall

Now gates which are outputting a 1 appear red and faults appear as a crossed out gate.

[michaelgarvie]

Please keep voting so that the most needed things get done. Just let me know of any feature requests not on the list so I can add them.

I'm currently working on the graph of computing effort. I'm not sure if to put this in the client or on the website...

[spaetz]

I would rather keep the client as lightweight as possible and put stats on the website. On the other hand, this does increase the load on your bandwidth (don't know if your institution has a max limit).

When thinking about improvements: I often run it in a non-graphical environment. While it just works, there are these awt related error messages. Is it possible to have a command line option that turns of the graphical window?

[michaelgarvie]

I know the error messages are slightly offending for someone used to debugging!! But they're totally harmless and waste negligble processing power (once every 2000000 * 60 clock cycles!)

I'll add it onto the ToDo list of the project which I intent to make public soon.

[michaelgarvie]

Answering Spaez's email:

Yeah the client is already 10Kb lightweight because of dynamical code download. This allows for automatic updating of the code. Since the project went live a lot of new code has been written and executed on the cluster for dealing with the multiple fault diagnosing circuits.

About the paper, the NASA/DoD one isn't strictly the latest latest.. There are a couple in publication right now:

http://www.cogs.susx.ac.uk/users/mmg20/files/iolts04_jig.pdf
and
http://www.cogs.susx.ac.uk/users/mmg20/files/iolts04_sc.pdf

They're also quite dense but hopefully the results table should show you how good this project's results are compared to the state of the art!!

[DrkNexus]

I just ran the GUI today for a little while and noticed that in its current form it isn't going to be much "fun" to watch at all. I should have kept it running to keep from sounding as stupid, but there are five inputs on this circuit however there are 28 outputs which make the GUI span well off my display, and I suspect most everybody else's as well.

Perhaps if I knew a little bit more about the math/GA the 'core' was doing then I could guess at how the GUI could be made to be more interesting.

Maybe just a status line - "Evolving at Grid x/y" - "Migrated to Grid x/y" - "Accepted migration from x/y" - "All your base are belong to us" - "Sharing clock cycles means caring" - whatever ... something that basicly tells me the decisions on a macro-level the client has made and why it has made them. Perhaps also one telling me about the status of the circut.

[Stephen_B]

Originally posted by miguelgarvie

http://www.cogs.susx.ac.uk/users/mmg20/files/iolts04_jig.pdf

They're also quite dense but hopefully the results table should show you how good this project's results are compared to the state of the art!!

I haven't been following the state-of-the-art in the GA world, but this seems to me to be an impressive accomplishment. I was not aware that GAs could be used dynamically in deployed circuits.

After all this effort with the FPGA, I hope that the power supply doesn't go ... :(

One typo I caught: on page 8 in the last paragraph, change "If the circuit were not in use then it could supplied inputs" to "If the circuit were not in use then it could be supplied inputs".

[michaelgarvie]

Thanks for your comments DrNexus, they will be taken into consideration when the GUI is redesigned. I would love to do this right now, but unfortunately I am a bit snowed under given that I have to write up a thesis... If only I were not alone! :help:

Stephen, thanks for your keen eye there, I'll get that typo sorted. :thumbs: