AForge.NET

  :: AForge.NET Framework :: Articles :: Forums ::

Using learn network

The forum is to discuss topics from different artificial intelligence areas, like neural networks, genetic algorithms, machine learning, etc.

Using learn network

Postby ganjawar » Tue Nov 11, 2014 1:55 pm

Hi/ I'm using example(Multilayer Feed Forward Neural Networks samples) for training network. Next, i'm save train network and open her in other Window, using load, for compute my input value, but my value(from example test file for training) does not coincide with the response
Exaple
1-1
2-1,5
3-2,5
4-4
5-6
6-8,5
7-11,5
8-15

this data for trainig, after training i'm save network, load and input value, example 6, but return a very different response even with errors,even though the network is trained to 0,3 error
ganjawar
 
Posts: 6
Joined: Tue Nov 11, 2014 1:48 pm

Re: Using learn network

Postby andrew.kirillov » Tue Nov 11, 2014 11:31 pm

Hello,

You can check values of neural network's weights (and threshold) before saving and after loading. If weight are the same, then compute shall produce same result. If not, then there is some issue with network saving/loading.
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3453
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: Using learn network

Postby ganjawar » Wed Nov 12, 2014 4:02 pm

weights and thresold norm.
я просто не пойму, может это из-за аппроксимации?я подал на обучение тестовый файл из примера.потом после обучения ввожу на вход одно из значений из текстового файла а выход очень не правильный.посмотрел по коду..там сначала числа к сигмоидальному виду приводяться и на выходе из него выволяться.мне что тоже так делать после обучения?или обучение с аппроксимацией не как не влияет на дальнейшее использование сети?
ganjawar
 
Posts: 6
Joined: Tue Nov 11, 2014 1:48 pm




Return to Artificial Intelligence

cron