Page 2 of 3

Re: Neural Network

Posted: Thu Aug 29, 2019 1:51 pm
by sphere
richmond62 wrote: Thu Aug 29, 2019 1:14 pm Thoughts . . .

http://www.tigabyte.com/hnfiles/neural.html
going to have a play with it, looks similair language, and even for Android

Re: Neural Network

Posted: Thu Aug 29, 2019 2:10 pm
by bogs
Well it was close, and I like that computer almost as much as I like HAL... :shock:
sphere wrote: Thu Aug 29, 2019 1:51 pm looks similair language, and even for Android
...but still no for 'nix :x which is kinda surprising, since RB does run on 'nix.

Re: Neural Network

Posted: Thu Aug 29, 2019 3:44 pm
by richmond62
HyperNEXT has been developed by one bloke, all by himself, and in my opinion, while being a very brave attempt, it is not up to LiveCode's standard. Having said that I think the LiveCode people should have a long look at HyperNEXT, and just possibly "pinch" some of the ideas there.

Re: Neural Network

Posted: Thu Aug 29, 2019 4:10 pm
by bogs
Yup, I've been watching it long enough to know its history, but as I say, I'm still surprised it hasn't been ported to 'nix yet.

Re: Neural Network

Posted: Thu Aug 29, 2019 7:38 pm
by sphere
on the website it says Linux development has ceased

Re: Neural Network

Posted: Thu Aug 29, 2019 8:05 pm
by richmond62
Malkom told me about 4 years ago that as he was working on his own with
almost no financial support he could not sustain Linux development, and, in
fact no Linux builds have ever got anywhere.

Re: Neural Network

Posted: Thu Aug 29, 2019 8:16 pm
by bogs
I think the reason for my surprise might not be understood. My assumption is that he was building this from RB, which I was using just prior to coming to Lc/Mc.

RB works on the same desktop platforms that Lc/Mc work on, so the reason I am 'surprised' is that the port shouldn't have required all that much to do having a complete IDE/language environment already, relatively speaking. I wonder if the difficulty he ran into was an issue with the libraries on 'nix, I know the older versions of RB that I have required some really older 32bit stuff that isn't even in the repositories anymore.

I have nothing but admiration for someone that takes on making a programming environment, whether alone or as part of a group.

Re: Neural Network

Posted: Fri Aug 30, 2019 7:47 am
by richmond62
I have e-mailed Malkom several times over the years and he has
always been very forthcoming and answered my questions . . .

info@tigabyte.com

Re: Neural Network

Posted: Fri Sep 06, 2019 9:35 am
by sphere
i'm just seeing these replies, as i don't get any emails from the forum it seems.
Thanks for your comments.

Re: Neural Network

Posted: Fri Sep 06, 2019 12:36 pm
by bogs
If you have it set to send you an email in the CP, then I would check your spam/trash box.

If they aren't sent there, maybe enable rss and hit the forum that way (but you will see everything posted).

Re: Neural Network

Posted: Fri Sep 06, 2019 1:19 pm
by sphere
ok, i thought it had something to do with the move.
thanks Bogs

Re: Neural Network

Posted: Fri Sep 06, 2019 1:38 pm
by bogs
That is also a possibility :D

Re: Neural Network

Posted: Wed Sep 11, 2019 1:45 pm
by richmond62
via graphs and text output
http://www.tigabyte.com/hnfiles/neural.html

Wondering about graph output in LiveCode.

https://livecode.com/livecode-widgets-the-line-chart/

Re: Neural Network

Posted: Mon Sep 16, 2019 7:13 pm
by richmond62
One of the "problems" of any attempt at developing a Neural Network is
that as it, supposedly, "learns" it must be able to modify its own code . . .

. . . now that would be asking a bit much in a standalone.

Re: Neural Network

Posted: Mon Sep 16, 2019 8:23 pm
by FourthWorld
richmond62 wrote: Mon Sep 16, 2019 7:13 pm One of the "problems" of any attempt at developing a Neural Network is
that as it, supposedly, "learns" it must be able to modify its own code . . .

. . . now that would be asking a bit much in a standalone.
Self-modifying code is easily done the way all modifications are done in all applications: put the part that needs to be saved as modified into a separate file.

With neural nets, though, it's not so much code that's modified, as weight tables.