Linux client

[J. Hart]

I too would be most interested in a linux client. If one is not yet
available, I would be most interested in creating and maintaining one if
there is any interest.

I have a fair bit of experience with this sort of thing, having worked
for quite a few years on the analysis code and tools for Tom Ray's
Network Tierra here in Japan.

J. Hart

[Michael Schmidt]

We're finalizing a simple c++ library and API right now, and a base
commandline client that can be built upon. This is a high priority and
should be released in the next week or two.

[J. Hart]

If there is anything I can do to be of assistance with work on the
client, I would be most happy to contribute my efforts in this regard.

J. Hart

eureqa...@googlegroups.com wrote:
>
> I have a fair bit of experience with this sort of thing, having
> worked
> for quite a few years on the analysis code and tools for Tom Ray's
> Network Tierra here in Japan.
>
> J. Hart
>
>
>

> Michael Schmidt <michael.dou...@gmail.com> Dec 05
> 01:26PM -0800

[J. Hart]

On the download page it states that one should contact the lab if one is
interested in the library and API as needed for development. I'd be
very interested in this, and would like to study the API if possible
when this is released.

With Thanks,

J. Hart

[J. Hart]

just a quick thought:

Would it be possible to have an option to export the "found" equations
in formats compatible with common symbolic algebra packages such as
Maxima, Maple or Mathematica ?

Regards,

J. Hart

[Michael Schmidt]

Great idea. If you double-click on the display equation you can get
its plain text, but we can also add special formatting for Mathematica
etc.

[J. Hart]

I am quite new at Eureqa, so I'm not so familiar with how "candidate
function" fitness is actually evaluated yet, but I'd like to express a
few thoughts on fitness functions and see what people think:

It might be useful to have Eureqa evaluate candidate solutions using a
fitness function which takes "candidate size" into account along with
its error characteristics. It may eliminate some of the more
computationally intensive candidates in favor of the less intensive ones
which may still be acceptably accurate.

Another possibility would be to track "evaluation rates" for individual
candidates. A fitness function which takes this rate into account might
also eliminate the more computationally intensive candidates in favor of
the less intensive, possibly simpler but still acceptably accurate
candidates.

Either of these might serve to considerably speed up a candidate search
by quicker elimination of "unnecessarily complex" candidates.

Regards,

J. Hart

[J. Hart]

I notice that the server has a fairly high cpu usage (~75 - 95%) even
when the search is paused. Why might this be ?

Note: I am presently running the "single client / single server" model
under Linux via Wine.

Regards,

J. Hart

[Hod Lipson]

If you select the equation and click on it in the window under "quick
statistics of solution" you can copy the formula in a standard text
expression and paste it into matlab etc. That's the best way to plot it in
3D and/or do further analysis.

--hod

--
Eureqa ( http://www.eureqa.com )
-------------------------------------------------
Unsubscribe: eureqa-group...@googlegroups.com
View Group: http://groups.google.com/group/eureqa-group?hl=en?hl=en