[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Gneuralnetwork] Adding on to Gneural Network
From: |
Karl Semich |
Subject: |
[Gneuralnetwork] Adding on to Gneural Network |
Date: |
Tue, 30 Aug 2022 09:18:29 -0400 |
Hey list,
Is anybody there? The archives show about 1 post a year. Please reply
if you're here!
I forked gneural network to https://github.com/xloem/gneuralknockoff .
I submitted a savannah request to open a new project, but haven't
heard back yet. I'm curious how gneural network become decommissioned,
or if maybe it was a mistake.
I'm working on helping stabilise the code, and have made progress. I
fixed a parser bug and got the old tests running. I'm spending a
little time now on the new nnet code which clearly has improved design
plans.
Although I've used and hacked on mainstream machine learning networks,
I don't have any formal education in neural networks, so I'm mostly
looking up concepts as I encounter them.
The nnet code reserves the first input for bias, but there appears to
be a bug where it attempts to read bias input values from the data.
I'm assuming that bias is intended only for trainable weights, not
input. I'm not sure what the appropriate solution to this is: should
the hardcoded bias input node be considered an input here, or not? Is
there ever a case where you would want to provide data for it?
Is anybody in contact with the original devs?
Karl Semich
Constant coda: I have a mental disability and am unlikely to bring a
work to production readiness on my own at this time, but I am
interested in contributing.
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [Gneuralnetwork] Adding on to Gneural Network,
Karl Semich <=