[Top][All Lists]

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

Re: [Gneuralnetwork] Eventual Role for command-line gneuralnetwork.

From: Mohammad Akhlaghi
Subject: Re: [Gneuralnetwork] Eventual Role for command-line gneuralnetwork.
Date: Mon, 10 Oct 2016 19:05:00 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.4.0

On 10/10/2016 06:54 PM, Ray Dillinger wrote:

On 10/09/2016 05:16 PM, Tom Acunzo wrote:
How does one start really? When you are new and backgrounds vary.

Do we create a relatable list of base NN examples with data and configuration 
to showcase GneuralNetwork features? A set of user case templates that can be 
revisited upon update...

Yes, we definitely need to have a list of NN examples with data and
configuration.  90%+ of the people out there learn from examples +
doc *far* better than they learn from doc alone.

But there may be a problem with that.  What you need to train
neural networks is massive data sets.  As installed we ought to
have a 'data' directory that contains MNIST, an image database,
and several other things.  But I have no idea whether Savannah
will be okay with a single project that occupies a very large
amount of disk space.  Or, when people come to download it, a
very large amount of bandwidth.

We may wind up having to put a torrent link in the doco for some
of the massive data sets, and then one or more of us can host the
torrent ourselves to make sure it never goes down.

One possible place you could keep and share large volumnes of data is Zenodo, which is a partner of CERN and even give the upload a DOI for citations:


reply via email to

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