gneuralnetwork
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gneuralnetwork] Welcome to Gneural Network


From: Jean Michel Sellier
Subject: Re: [Gneuralnetwork] Welcome to Gneural Network
Date: Wed, 16 Mar 2016 09:25:03 +0100

Yes, because for tasks like this one (i.e. defining a neural network and its tasks) one doesn't need too many commands so it's pretty easy and effective.

JM


2016-03-16 8:31 GMT+01:00 Nala Ginrut <address@hidden>:
Oops, you write yet another interpreter?

On Wed, 2016-03-16 at 08:25 +0100, Jean Michel Sellier wrote:
> The Guile project sounds interesting but the thing is that my
> scripting language is almost done.. which is a good thing because we
> don't want to wait too long, right?
>
>
> JM
>
>
>
> 2016-03-16 7:49 GMT+01:00 Nala Ginrut <address@hidden>:
>         Hi Jean!
>
>         On Tue, 2016-03-15 at 17:03 +0100, Jean Michel Sellier wrote:
>         > Very soon, the following features will be released (I am in
>         the process of
>         > rewriting some code that I already developed and tested):
>         >
>         > - a scripting language, so that the user can define and run
>         a neural
>         > networks without even knowing how to program,
>
>         As a official GNU project, it's naturally to use GNU Guile as
>         extension
>         scripting platform. The main language is Scheme, but a good
>         news is that
>         the Lua frontend which I'm working on maybe released in this
>         year. It
>         means that once you use GNU Guile API for extension, you have
>         at least
>         Scheme and Lua as scripting languages. And the modules written
>         in Scheme
>         or Lua could be applied interactively.
>         One effort, more results. Python? Maybe later. ;-)
>         To whom may concern, here's the project:
>         https://github.com/NalaGinrut/guile-lua-rebirth
>
>
>         > - a repository will be created to make interactions easier,
>
>         Savannah is definitely the main choice, but Savannah supports
>         many VCS
>         system(CVS/hg/git/svn...), I recommend we just use Git. It is
>         necessary
>         to apply a repo on Savannah with manually review by other GNU
>         folks.
>         Please don't hesitate to ask if you encounter any problem.
>
>         Besides, personally I prefer github as a mirror. For better
>         maintain,
>         you may need a proper config in local git repo to push the
>         commits to
>         github/savannah both, automatically.
>
>         I'm using both mailing-list and github issues to communicate
>         with users.
>         But it depends on you to use mailing-list only.
>
>         > - recurrent networks, in particular LSTM neurons,
>         > - an automatic algorithm which select the best network
>         topology given a
>         > specific problem,
>
>         I'm confusing here, do you mean an abstract mathematical
>         network
>         topology? or detecting the fastest physical network node for
>         connection
>         speed?
>
>         > - parallelization of the training process for multi-core
>         CPUs and GPUs,
>
>         I'm not expertise on GPU, but if you want to use CUDA for
>         parallel, it
>         seems impossible on Nouveau at present.
>         But someone is working on it:
>         https://www.phoronix.com/scan.php?page=news_item&px=NVIDIA-Nouveau-CUDA-Support
>
>         Alternatively, we may choose OpenCL, but Nouveau seems not
>         prepared yet:
>         https://nouveau.freedesktop.org/wiki/FeatureMatrix/
>
>         But we still have some hope:
>         https://fosdem.org/2016/schedule/event/nv_compute/
>
>         Anyway, we have two choices:
>         1. Using propriety GPU driver temporally, not good for GNU
>         ethic, but
>         practical.  Waiting for the Nouveau.
>
>         2. Using CPU for it, and waiting for the Nouveau. The problem
>         is the
>         time, Gneural-network folks may lose their patient.
>
>
>
>         I may get wrong, comments are welcome.
>
>         Best regards.
>
>
>         --
>         GNU Powered it
>         GPL Protected it
>         GOD Blessed it
>
>         HFG - NalaGinrut
>
>         Fingerprint 77F8 7D5B B425 E505 535E  C365 9182 7E28 EE78 E925
>
>
>
>
>
>
> --
> Jean Michel Sellier, Ph.D., M.Sc.
>
>
> www.nano-archimedes.com
>
> www.gnu.org/software/archimedes
>
>
> www.gnu.org/software/nano-archimedes
>
>
>
> www.gnu.org/software/gneuralnetwork
>
>
> http://en.wikipedia.org/wiki/GNU_Archimedes
>
>

--
GNU Powered it
GPL Protected it
GOD Blessed it

HFG - NalaGinrut

Fingerprint 77F8 7D5B B425 E505 535E  C365 9182 7E28 EE78 E925





--

reply via email to

[Prev in Thread] Current Thread [Next in Thread]