Re: HELP WITH THE EDELMAN NETWORK

Hi Juan

That 11 years is not the hidden neurons.It is time series data of 11
years. So take 2 or 3 hidden layers and make the 1st layers 8,2nd
layer 6 and 3rd layer 4 neurons etc.
Increase the no of epocs to 10000 and net.trainParam.goal make it
1e-6. change traingdx to trainlm.

train the network. if you will see that performance is not good then
increase the no of neurons in the hidden layer and change the
transfer function tansig to purelin.

Regards;
Ambarish

Juan Carlos Morales Aragón wrote:
>
>
> Hi everybody
>
> I am trying to train a Edelman network with the Neural Networks
> Tool
> box, with the the classical date base of the sunspots number, this
> set was normalized, the set was divided in two parts Y: is input
> and
> U is output
>
> Y=[0.0262 0.0575
> 0.0837 0.1203 0.1883 0.3033 0.1517 0.1046 0.0523 0.0418 0.0157 0 0
> 0.0105
>
0.0575 0.3818 0.2458 0.1831 0.0575 0.0262 0.0837 0.1778 0.3661 0.423
> 6
>
0.5805 0.5282 0.3818 0.2092 0.1046 0.0837 0.0262 0.0575 0.1151 0.209
> 2
> 0.3138 0.4231 0.4362 0.2495 0.25
>
0.1606 0.0638 0.0502 0.0534 0.17 0.2489 0.2824 0.329 0.4493 0.3201 0
> .2
>
359 0.1904 0.1093 0.0596 0.1977 0.3651 0.5549 0.5272 0.4268 0.3478 0
> .1
>
82 0.16 0.0366 0.1036 0.4838 0.8075 0.6585 0.4435 0.3562 0.2014 0.11
> 92
> 0.0534 0.126 0.4336 0.6904 0.6846
>
0.6177 0.4702 0.3483 0.3138 0.2453 0.2144 0.1114 0.0837 0.0335 0.021
> 4
>
0.0356 0.0758 0.1778 0.2354 0.2254 0.2484 0.2207 0.147 0.0528 0.0424
> 0
> .0131 0 0.0073
> 0.0262 0.0638 0.0727 0.1851 0.2395 0.215 0.1574 0.125 0.0816
> 0.0345 0.0209 0.0094 0.0445 0.0868 0.1898 0.2594 0.3358 0.3504
>
0.3708 0.25 0.1438 0.0445 0.069 0.2976 0.6354 0.7233 0.5397 0.4482 0
> .3
> 379 0.1919 0.1266 0.056 0.0785 0.2097
> 0.3216 0.5152 0.6522 0.5036 0.3483 0.3373 0.2829]
> U=[0.204 0.1077
> 0.035 0.0225 0.1187 0.2866 0.4906 0.501 0.4038 0.3091 0.2301 0.2458
> 0.
> 1595 0.0853
> 0.0382 0.1966 0.387 0.727 0.5816 0.5314 0.3462 0.2338 0.0889 0.0591
> 0.
> 0649 0.0178 0.0314
>
0.1689 0.284 0.3122 0.3332 0.3321 0.273 0.1328 0.0685 0.0356 0.033 0
> .0
> 371 0.1862 0.3818 0.4451
>
0.4079 0.3347 0.2186 0.137 0.1396 0.0633 0.0497 0.0141 0.0262 0.1276
> 0
> .2197 0.3321 0.2814 0.3243
>
0.2537 0.2296 0.0973 0.0298 0.0188 0.0073 0.0502 0.2479 0.2986 0.543
> 4
> 0.4215 0.3326 0.1966 0.2537
>
0.2296 0.0973 0.0298 0.0188 0.0073 0.0502 0.2479 0.2986 0.5434 0.421
> 5
> 0.3326 0.1966 0.1365
>
0.0743 0.0303 0.0873 0.2317 0.3342 0.3609 0.4069 0.3394 0.1867 0.110
> 9
> 0.0581 0.0298 0.0455 0.1888
> 0.4168 0.5983 0.5732 0.4644 0.3546 0.2484 0.16 0.0853 0.0502 0.1736
> 0.
> 4843 0.7929 0.7128
> 0.7045 0.4388 0.363 0.1647 0.0727 0.023 0.1987 0.7411 0.9947 0.9665
> 0.
> 8316 0.5873 0.2819 0.1961
>
0.1459 0.0534 0.079 0.2458 0.4906 0.5539 0.5518 0.5465 0.3483 0.3603
> 0
> .1987 0.1804 0.0811 0.0659 0.1428 0.4838]
>
> In the literature we can to read about of the regularity of this
> phenomena is almost 11 year, for them I thing that the hidden layer
> must have 11 neurons, the other layers are: one neuron of input and
> one neuron output. I am using the follow sequence of the lines:
>
> Pseq = con2seq(Y);
> Tseq = con2seq(U);
> R = 1;% 1 input element
> S2 = 1;% 1 layer 2 output neuron
> S1 = 11;% 11 recurrent neurons in the first layer
> net = newelm([0 1],[S1 S2],{'tansig','purelin'},'traingdx');
> net.trainParam.epochs = 6000;
> net.trainParam.goal = 0.001
> [net,tr] = train(net,Pseq,Tseq);
>
> With this lines the error was in the interval (0.3-0.4) for them
> the
> values forecasting to U Set has been very bad.
>
> I did some things to improve the perform of the network, for
> example:
> 1.-To increased the number of the recurrent neurons for example 25
> 2.- To change the learning constant and momentum constant
> 3.- To give more epochs for example 3000-12000.
> 4.- To change of transfer function from hidden layer to output
> layer,
> the default function is "purelin", this function was changed to
> "tansig"
>
> If some person would give me any key to improved my network, I will
> very grateful with this help and information.
>
> Sincerely
> Juan Carlos Aragón
.

Relevant Pages

  • Re: Training code differences for premnmx and mapminmax
    ... % preprocess input data P so that minimum is -1 and maximum is 1 in input value ... % tansig transfer function as the transfer function in the hidden layer and ... % traingdx as the network training function ...
    (comp.soft-sys.matlab)
  • Re: Training code differences for premnmx and mapminmax
    ... % preprocess input data P so that minimum is -1 and maximum is 1 in input value ... % tansig transfer function as the transfer function in the hidden layer and ... % traingdx as the network training function ...
    (comp.soft-sys.matlab)
  • Training code differences for premnmx and mapminmax
    ... % preprocess input data P so that minimum is -1 and maximum is 1 in input value ... % tansig transfer function as the transfer function in the hidden layer and ... % traingdx as the network training function ...
    (comp.soft-sys.matlab)
  • HELP WITH THE EDELMAN NETWORK
    ... I am trying to train a Edelman network with the Neural Networks Tool ... 1.-To increased the number of the recurrent neurons for example 25 ... 2.- To change the learning constant and momentum constant ... 4.- To change of transfer function from hidden layer to output layer, ...
    (comp.soft-sys.matlab)
  • Re: Economy of ideas
    ... multiple neurons. ... there are systems at work regulating brain activity. ... Either stimulus signals would die out quickly in the network, ... concept of "economy of ideas". ...
    (comp.ai.philosophy)