Path: news.mathworks.com!not-for-mail
From: <HIDDEN>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Question about number of layers in ANN
Date: Thu, 29 Oct 2009 08:37:02 +0000 (UTC)
Organization: The MathWorks, Inc.
Lines: 16
Message-ID: <hcbk7e$2a2$1@fred.mathworks.com>
References: <hcbiur$5g7$1@fred.mathworks.com> <hcbjf6$a7o$1@fred.mathworks.com>
Reply-To: <HIDDEN>
NNTP-Posting-Host: webapp-02-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1256805422 2370 172.30.248.37 (29 Oct 2009 08:37:02 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Thu, 29 Oct 2009 08:37:02 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 1669376
Xref: news.mathworks.com comp.soft-sys.matlab:580867


Thank you, Gavrilo,
But could you please provide some more detail about the relationship between the input and the number of layers?


"Gavrilo Bozovic" <gavrilo.dot.bozovic@gmail.dot.ch> wrote in message <hcbjf6$a7o$1@fred.mathworks.com>...
> "Kloser " <kcfcheung@gmail.com> wrote in message <hcbiur$5g7$1@fred.mathworks.com>...
> > Dear All
> > I want to know, in Neural Networks, whether more layers will promise higher accuracy for classification?
> > I have tried using two networks to classify some images into classes. There are 100 variables in each sample, meaning that there are 100 input nodes. And for the Networks, The 1-layer network have 60 output nodes, and the 2-layers network have 20 hidden nodes in the hidden layers and 60 output nodes. 
> > But I found that the accuracy in the 1-layer network is higher than that in the 2-layer. I expected that the 2-layer may have higher accuracy because there are more parameters, so I am wondering whether there are bugs in my implementation. But I can't find any bugs up to now.
> > So can anyone tell me if that is my wrong expectation or other reasons so I have this result? And what is the relationship between the numbers of layers and the accuracy? How should I judge the optimal numbers of layers for neural networks?
> > Thank you
> 
> If you have more layers, you will have a higher accuracy for classification, but only if you have an extended training set.
> 
> If you can't train your neural network on enough inputs, adding layers won't help, on the contrary.