I guess I’m not clear on what the inputs and outputs of the RNN would be. Would I send in various JSON documents with a YES/NO flag? Or would I send in each JSON document with incrementally updated schema? Obviously I would get the schema as output. How would I determine an error for back propagation? Sorry, my knowledge of RNNs is limited. I am trying to get a library, but so far I see netCDF and I have to convert my JSON documents to netCDF and download a bunch of packages. Also, my JSON documents are nested and hierarchical in nature.
John > On Nov 4, 2015, at 10:27 PM, Matthew Retchin <mretc...@gmail.com> wrote: > > If you have a ton of examples to train on, that makes a recurrent neural > network even more attractive as (part of) a solution, doesn't it? > > On Wed, Nov 4, 2015 at 11:22 PM, John Carlson <yottz...@gmail.com > <mailto:yottz...@gmail.com>> wrote: > Sorry, I said that wrong. I am trying to generate a single JSON schema from > thousands of JSON files. I don’t want a different JSON schema per JSON > document. I think that might make a difference as to what technology to use. > > Thanks, > > John > >> On Nov 4, 2015, at 9:36 PM, David Barbour <dmbarb...@gmail.com >> <mailto:dmbarb...@gmail.com>> wrote: >> >> You could probably repurpose a recurrent neural network to compute a likely >> schema in polynomial time. >> >> >> On Wed, Nov 4, 2015, 8:03 PM John Carlson <yottz...@gmail.com >> <mailto:yottz...@gmail.com>> wrote: >> I am familiar with Grammex, but I want wondering if there’s been any more >> effort to create syntax by incremental example, given that validation tools >> are present (online, either getting feedback from a compiler, or ideally >> from an IDE with incremental compiling and code completion). I think from >> reading about reverse engineering protocols that this might be polynomial >> time, whereas, without a validator (offline), it’s NP-complete? Or is that >> only for regular grammars? What about “syntax-free” languages, like machine >> code? Is this where the idea that it’s NP-complete came from? I’m actually >> interested in creating JSON schemas from thousands of example JSON files. I >> want something more detailed than the JSON specification, with domain >> validation. >> John >> _______________________________________________ >> Fonc mailing list >> Fonc@mailman.vpri.org <mailto:Fonc@mailman.vpri.org> >> http://mailman.vpri.org/mailman/listinfo/fonc_mailman.vpri.org >> <http://mailman.vpri.org/mailman/listinfo/fonc_mailman.vpri.org> > > > _______________________________________________ > Fonc mailing list > Fonc@mailman.vpri.org <mailto:Fonc@mailman.vpri.org> > http://mailman.vpri.org/mailman/listinfo/fonc_mailman.vpri.org > <http://mailman.vpri.org/mailman/listinfo/fonc_mailman.vpri.org> > >
_______________________________________________ Fonc mailing list Fonc@mailman.vpri.org http://mailman.vpri.org/mailman/listinfo/fonc_mailman.vpri.org
