Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

Thread Subject:
Training multiple data for a single feedforwardnet

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 17 Oct, 2012 20:08:17

Message: 1 of 20

I'm building a feedforwardnet like this:

(..)
P=[V';ia';w'];
T=[tq'];
net=feedforwardnet([5 25],'trainbr');
(..)

How could i train this neural net for more then one group '[V';ia';w']' ? How is the matlab structure to perform this kind of training?

Note that 'P' in this case is a 10006x3 matrix that i extract from a motor model.

Thanks in advance.

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 19 Oct, 2012 10:44:08

Message: 2 of 20

"Carlos Aragon" wrote in message <k5n37h$ier$1@newscl01ah.mathworks.com>...
> I'm building a feedforwardnet like this:
>
> (..)
> P=[V';ia';w'];
> T=[tq'];
> net=feedforwardnet([5 25],'trainbr');

One hidden layer with H nodes is sufficient. Try to minimize H by trial and error. Start with 10 small values of H and Ntrials = 10 of random initial weights for each value of H. For examples search using some of the following keywords:

heath close clear Ntrials Neq Nw

> (..)
>
> How could i train this neural net for more then one group '[V';ia';w']' ? How is the matlab structure to perform this kind of training?

For I dimensional inputs and O-dimensional outputs

[ I N ] = size(input)
[ O N] = size(target)

yielding

Neq = N*O training equations for estimating

Nw = (I+1)*H+(H+1)*O weights.

Neq >= Nw when

H <= (Neq-O)/(I+O+1)
 
> Note that 'P' in this case is a 10006x3 matrix that i extract from a motor model.

That is a large number of samples. You can probably use a simple 0.34/0.33/0.33 trn/val/tst random data split, net.divideFcn = 'dividetrain' and 'trainlm'.

Hope this helps.


Greg

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 20 Oct, 2012 22:41:08

Message: 3 of 20

"Carlos Aragon" wrote in message <k5n37h$ier$1@newscl01ah.mathworks.com>...
> I'm building a feedforwardnet like this:
>
> (..)
> P=[V';ia';w'];
> T=[tq'];
> net=feedforwardnet([5 25],'trainbr');
> (..)
>
> How could i train this neural net for more then one group '[V';ia';w']' ? How is the matlab structure to perform this kind of training?
>
> Note that 'P' in this case is a 10006x3 matrix that i extract from a motor model.
 
The issue here is that after training with set1, the weights will forget set1
while they are learning set 2. There are a variety of ways to mitigate forgetting.

The best is to use a modication of NEWRB that allows the input of an initial
hidden layer. Then

1. After training with set1, use those weights as initial weights for training with set2 + set1.

2. After training with set1, use those weights as initial weights for training with set2 and a "characteristic subset" of set1. The drawback is how to define that characteristic.

The reason this works is that each hidden node basis function has local region of influence and a 1-to-1 correspondence with a previous worst classified training vector.

Hope this helps.

Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 29 Oct, 2012 18:29:37

Message: 4 of 20

Greg, thanks in advance. You're helping a lot!

You said:

(..)

 The best is to use a modication of NEWRB that allows the input of an initial
> hidden layer. Then
>
> 1. After training with set1, use those weights as initial weights for training with set2 + set1.
>
> 2. After training with set1, use those weights as initial weights for training with set2 and a "characteristic subset" of set1. The drawback is how to define that characteristic.
>
> The reason this works is that each hidden node basis function has local region of influence and a 1-to-1 correspondence with a previous worst classified training vector.

(...)

I'm facing problems to perform this action on matlab. Is there any automated way there i can record set1 and then use it to train a set2? How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.

Thanks!!

"Greg Heath" <heath@alumni.brown.edu> wrote in message <k5v9a4$pj6$1@newscl01ah.mathworks.com>...
> "Carlos Aragon" wrote in message <k5n37h$ier$1@newscl01ah.mathworks.com>...
> > I'm building a feedforwardnet like this:
> >
> > (..)
> > P=[V';ia';w'];
> > T=[tq'];
> > net=feedforwardnet([5 25],'trainbr');
> > (..)
> >
> > How could i train this neural net for more then one group '[V';ia';w']' ? How is the matlab structure to perform this kind of training?
> >
> > Note that 'P' in this case is a 10006x3 matrix that i extract from a motor model.
>
> The issue here is that after training with set1, the weights will forget set1
> while they are learning set 2. There are a variety of ways to mitigate forgetting.
>
> The best is to use a modication of NEWRB that allows the input of an initial
> hidden layer. Then
>
> 1. After training with set1, use those weights as initial weights for training with set2 + set1.
>
> 2. After training with set1, use those weights as initial weights for training with set2 and a "characteristic subset" of set1. The drawback is how to define that characteristic.
>
> The reason this works is that each hidden node basis function has local region of influence and a 1-to-1 correspondence with a previous worst classified training vector.
>
> Hope this helps.
>
> Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 29 Oct, 2012 23:10:08

Message: 5 of 20

Is there any way that i could present, for example, 14 data set for training and then validate for just another 3 ? How can i do it?

Thanks!.

Carlos.

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 31 Oct, 2012 15:42:08

Message: 6 of 20

PLEASE, PLEASE DO NOT TOP POST!!!

"Carlos Aragon" wrote in message <k6mhuh$etk$1@newscl01ah.mathworks.com>...
> Greg, thanks in advance. You're helping a lot!
>
> You said:
>
> (..)
>
> The best is to use a modication of NEWRB that allows the input of an initial
> > hidden layer. Then
> >
> > 1. After training with set1, use those weights as initial weights for training with set2 + set1.

or, if you are lucky
 
> > 2. After training with set1, use those weights as initial weights for training with set2 and a "characteristic subset" of set1. The drawback is how to define that characteristic.
> >
> > The reason this works is that each hidden node basis function has local region of influence and a 1-to-1 correspondence with a previous worst classified training vector.
>
> (...)
>
> I'm facing problems to perform this action on matlab.

That statement is absolutely useless. I thought you wanted my help.

> Is there any automated way there i can record set1 and then use it to train a set2?

I have no idea what the second part of that statement means.

>How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.

Then simultaneously train on samples or characteristic exemplars from all 14.

If all of the data is not available at once, do it in stages.

What do you not understand about that?

Hope that is clear.

 Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 1 Nov, 2012 22:28:08

Message: 7 of 20

"Greg Heath" <heath@alumni.brown.edu> wrote in message <k6rgsg$sp3$1@newscl01ah.mathworks.com>...
> PLEASE, PLEASE DO NOT TOP POST!!!
>
> "Carlos Aragon" wrote in message <k6mhuh$etk$1@newscl01ah.mathworks.com>...
> > Greg, thanks in advance. You're helping a lot!
> >
> > You said:
> >
> > (..)
> >
> > The best is to use a modication of NEWRB that allows the input of an initial
> > > hidden layer. Then
> > >
> > > 1. After training with set1, use those weights as initial weights for training with set2 + set1.
>
> or, if you are lucky
>
> > > 2. After training with set1, use those weights as initial weights for training with set2 and a "characteristic subset" of set1. The drawback is how to define that characteristic.
> > >
> > > The reason this works is that each hidden node basis function has local region of influence and a 1-to-1 correspondence with a previous worst classified training vector.
> >
> > (...)
> >
> > I'm facing problems to perform this action on matlab.
>
> That statement is absolutely useless. I thought you wanted my help.
>
> > Is there any automated way there i can record set1 and then use it to train a set2?
>
> I have no idea what the second part of that statement means.
>
> >How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.
>
> Then simultaneously train on samples or characteristic exemplars from all 14.

> If all of the data is not available at once, do it in stages.

I have all the training and test data, but i dont know how could i do to train 14 training vectors and then validate it with just 1 set to check if the neural net is generalizing well.

> What do you not understand about that?
>
> Hope that is clear.
>
> Greg

Tying to be clear about wat i'm doing. here is the code:

ia=linear_train_1(1:5001,4);
w=linear_train_1(1:5001,5);
tq=linear_train_1(1:5001,2);
T1=[198:0.000799840032:202]; % Voltage is between 198V and 202V
iateste1=ia_lin_1(1:5001,4);
wteste1=ia_lin_1(1:5001,5);
P=[T1;ia';w']; % This is the training vector that in this case, trains just 1 set of data.
T=[tq']; I want my neural net to recognize 14 samples of [T1;ia;w']. T1 is fix but 'ia'' and 'w'' varies according to the load equation i'm changing on my motor model. The question is How could i train it to recognize those 14 samples? If i make 'Ia'' and 'w'' a matrix of 14 different currents and speed, this neural net do not allow me to test a simple vector like is below.

net=feedforwardnet([5 25],'trainbr');
net.trainParam.goal = 0.005; %error
net.trainParam.epochs = 2000;
net=train(net,P,T);
P1=[T1;iateste1';wteste1'];
Y = sim(net,P1);

As you can see, i'm not an expert on this ... i imagine if you could help me build this process of train and validate. Thanks a lot for your help!

Carlos.

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 2 Nov, 2012 03:55:08

Message: 8 of 20

"Carlos Aragon" wrote in message <k6ut1o$1mo$1@newscl01ah.mathworks.com>...
> "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6rgsg$sp3$1@newscl01ah.mathworks.com>...
> > PLEASE, PLEASE DO NOT TOP POST!!!
> >
> > "Carlos Aragon" wrote in message <k6mhuh$etk$1@newscl01ah.mathworks.com>...
> > > Greg, thanks in advance. You're helping a lot!
> > >
> > > You said:
> > >
> > > (..)
> > >
> > > The best is to use a modication of NEWRB that allows the input of an initial
> > > > hidden layer. Then
> > > >
> > > > 1. After training with set1, use those weights as initial weights for training with set2 + set1.
> >
> > or, if you are lucky
> >
> > > > 2. After training with set1, use those weights as initial weights for training with set2 and a
"characteristic subset" of set1. The drawback is how to define that characteristic.
> > > >
> > > > The reason this works is that each hidden node basis function has local region of influence
and a 1-to-1 correspondence with a previous worst classified training vector.
> > >
> > > (...)
> > >
> > > I'm facing problems to perform this action on matlab.
> >
> > That statement is absolutely useless. I thought you wanted my help.
> >
> > > Is there any automated way there i can record set1 and then use it to train a set2?
> >
> > I have no idea what the second part of that statement means.
> >
> > >How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.
> >
> > Then simultaneously train on samples or characteristic exemplars from all 14.
>
> > If all of the data is not available at once, do it in stages.
>
> I have all the training and test data, but i dont know how could i do to train 14
training vectors and then validate it with just 1 set to check if the neural net is generalizing well.

Not even close. See below.
 
> Tying to be clear about wat i'm doing. here is the code:
>
> ia=linear_train_1(1:5001,4);
> w=linear_train_1(1:5001,5);
> tq=linear_train_1(1:5001,2);
> T1=[198:0.000799840032:202]; % Voltage is between 198V and 202V
> iateste1=ia_lin_1(1:5001,4);
> wteste1=ia_lin_1(1:5001,5);

Seems finely spaced. Do you reallyy need this much data? See below.

> P=[T1;ia';w']; % This is the training vector that in this case, trains just 1 set of data.
> T=[tq']; I want my neural net to recognize 14 samples of [T1;ia;w']. T1 is fix but
'ia'' and 'w'' varies according to the load equation i'm changing on my motor model. The
question is How could i train it to recognize those 14 samples? If i make 'Ia'' and 'w'' a
matrix of 14 different currents and speed, this neural net do not allow me to test a simple
>vector like is below

You need to test matrices not single vectors..
 
> net=feedforwardnet([5 25],'trainbr');

Why 2 hidden layers??? Why H =25 ?? Why 'trainbr?

> net.trainParam.goal = 0.005; %error

Why?

> net.trainParam.epochs = 2000;

Why?

> net=train(net,P,T);

Performance evaluation??

[ net tr ] = ...
MSEtrn = ?
MSEval =?
MSEtst = ?

Otherwise, how do you obtain separate tr/val/tst results.

> P1=[T1;iateste1';wteste1'];
> Y = sim(net,P1);
>
> As you can see, i'm not an expert on this ... i imagine if you could help me build this
    process of train and validate. Thanks a lot for your help!

This is post No. 8 of this thread and you don't seem to be any further along than you were
at the first post. So, let's start again

1. What is a motor model?
2. What is a motor load?
3. What are V, ia, w and tq ?
4.What are the corresponding correlation coefficients?
5. What , exactly, are the differences between the 14 data sets?
6. Have you plotted the output to determine how much sample spacing is
needed to adequately characterize it?
7. Given that spacing, how much data is needed for that characterization?
8. Your first post mentions 10,006 measurements but later you use 5,0001.
Is that for each of the 14 data sets?
9. As I stated before
   1. Only 1 hidden layer is necessary
   2. If you have 14 scenarios that you want to characterize with one net:
       a. Take 6 and 7 into consideration and combine samples of all 14 into
           multiple mixed subsets.
       b. Since you have a large data set, Train/Validate and Test with a
           0.34/0.33/0.33 data split.
       c. Use one or more data sets, as many defaults as possible, and vary
           H to find the minimum acceptable value.

This should give you a solid start.

Hope this helps.

Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 3 Nov, 2012 21:33:14

Message: 9 of 20

"Greg Heath" <heath@alumni.brown.edu> wrote in message <k6vg6s$36r$1@newscl01ah.mathworks.com>...
> "Carlos Aragon" wrote in message <k6ut1o$1mo$1@newscl01ah.mathworks.com>...
> > "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6rgsg$sp3$1@newscl01ah.mathworks.com>...
> > > PLEASE, PLEASE DO NOT TOP POST!!!
> > >
> > > "Carlos Aragon" wrote in message <k6mhuh$etk$1@newscl01ah.mathworks.com>...
> > > > Greg, thanks in advance. You're helping a lot!
> > > >
> > > > You said:
> > > >
> > > > (..)
> > > >
> > > > The best is to use a modication of NEWRB that allows the input of an initial
> > > > > hidden layer. Then
> > > > >
> > > > > 1. After training with set1, use those weights as initial weights for training with set2 + set1.
> > >
> > > or, if you are lucky
> > >
> > > > > 2. After training with set1, use those weights as initial weights for training with set2 and a
> "characteristic subset" of set1. The drawback is how to define that characteristic.
> > > > >
> > > > > The reason this works is that each hidden node basis function has local region of influence
> and a 1-to-1 correspondence with a previous worst classified training vector.
> > > >
> > > > (...)
> > > >
> > > > I'm facing problems to perform this action on matlab.
> > >
> > > That statement is absolutely useless. I thought you wanted my help.
> > >
> > > > Is there any automated way there i can record set1 and then use it to train a set2?
> > >
> > > I have no idea what the second part of that statement means.
> > >
> > > >How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.
> > >
> > > Then simultaneously train on samples or characteristic exemplars from all 14.
> >
> > > If all of the data is not available at once, do it in stages.
> >
> > I have all the training and test data, but i dont know how could i do to train 14
> training vectors and then validate it with just 1 set to check if the neural net is generalizing well.
>
> Not even close. See below.
>
> > Tying to be clear about wat i'm doing. here is the code:
> >
> > ia=linear_train_1(1:5001,4);
> > w=linear_train_1(1:5001,5);
> > tq=linear_train_1(1:5001,2);
> > T1=[198:0.000799840032:202]; % Voltage is between 198V and 202V
> > iateste1=ia_lin_1(1:5001,4);
> > wteste1=ia_lin_1(1:5001,5);
>
> Seems finely spaced. Do you reallyy need this much data? See below.
>
> > P=[T1;ia';w']; % This is the training vector that in this case, trains just 1 set of data.
> > T=[tq']; I want my neural net to recognize 14 samples of [T1;ia;w']. T1 is fix but
> 'ia'' and 'w'' varies according to the load equation i'm changing on my motor model. The
> question is How could i train it to recognize those 14 samples? If i make 'Ia'' and 'w'' a
> matrix of 14 different currents and speed, this neural net do not allow me to test a simple
> >vector like is below
>
> You need to test matrices not single vectors..
>
> > net=feedforwardnet([5 25],'trainbr');
>
> Why 2 hidden layers??? Why H =25 ?? Why 'trainbr?
>trainbr: i have to use bayesien-regulation
> > net.trainParam.goal = 0.005; %error
>
> Why?
defined goal.
> > net.trainParam.epochs = 2000;
>
> Why?
maximum training epochs that i want, if not reached the trainparam.goal.
> > net=train(net,P,T);
>
> Performance evaluation??
>
> [ net tr ] = ...
> MSEtrn = ?
> MSEval =?
> MSEtst = ?

I dont understand what it means.

> Otherwise, how do you obtain separate tr/val/tst results.
>
> > P1=[T1;iateste1';wteste1'];
> > Y = sim(net,P1);
> >
> > As you can see, i'm not an expert on this ... i imagine if you could help me build this
> process of train and validate. Thanks a lot for your help!
>
> This is post No. 8 of this thread and you don't seem to be any further along than you were
> at the first post. So, let's start again

It's a dificult task to explain. The goal of this thread is (code by code) determine how to train different sets of [V;Ia;w] defined above, so that my neural net will recognize those 14 datas.
 
> 1. What is a motor model?
Simulink-SimPowerSystems

There's an induction motor machine model. Resuming, i'm extracting data from this model associated with other procediment that does not matter here..

> 2. What is a motor load?

A motor is a device that converts electrical energy into mechanical energy to act upon a mechanical load. The burden placed on the motor due to this mechanical activity is referred to as the motor load.

> 3. What are V, ia, w and tq ?
V -> Voltage
ia -> Current on phase 'a'
w-> motor speed
tq-> it's the load generated according to the type of burden used to train.

> 4.What are the corresponding correlation coefficients?
I think that does'nt matter too

> 5. What , exactly, are the differences between the 14 data sets?
Defined what load is, the difference between the 14 data sets is the type with burden i'm using on the motor.

> 6. Have you plotted the output to determine how much sample spacing is
> needed to adequately characterize it?
5000 datas is enough to have values from the transitory state to steady state.

> 7. Given that spacing, how much data is needed for that characterization?
6.
> 8. Your first post mentions 10,006 measurements but later you use 5,0001.
Yes. I've cut unnecessary data.

> Is that for each of the 14 data sets?
one data set is a value of V ; Ia; W. Only 'V' is fix. Ia nd W varies in each of the 14 data sets. There are 5001 values of Ia and 5001 Values of W as there are 5001 values of fixed V (voltage)

> 9. As I stated before
> 1. Only 1 hidden layer is necessary
> 2. If you have 14 scenarios that you want to characterize with one net:
> a. Take 6 and 7 into consideration and combine samples of all 14 into
> multiple mixed subsets.
> b. Since you have a large data set, Train/Validate and Test with a
> 0.34/0.33/0.33 data split.
> c. Use one or more data sets, as many defaults as possible, and vary
> H to find the minimum acceptable value.
>
> This should give you a solid start.
>
> Hope this helps.
>
> Greg

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 3 Nov, 2012 23:40:16

Message: 10 of 20

"Carlos Aragon" wrote in message <k742iq$il8$1@newscl01ah.mathworks.com>...
> "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6vg6s$36r$1@newscl01ah.mathworks.com>...
> > "Carlos Aragon" wrote in message <k6ut1o$1mo$1@newscl01ah.mathworks.com>...
> > > "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6rgsg$sp3$1@newscl01ah.mathworks.com>...
> > > > PLEASE, PLEASE DO NOT TOP POST!!!
> > > >
> > > > > The best is to use a modication of NEWRB that allows the input of an initial
> > > > > > hidden layer. Then
> > > > > >
> > > > > > 1. After training with set1, use those weights as initial weights for training with set2 + set1.
> > > >
> > > > or, if you are lucky
> > > >
> > > > > > 2. After training with set1, use those weights as initial weights for training with set2 and a
> > "characteristic subset" of set1. The drawback is how to define that characteristic.
> > > > > >
> > > > > > The reason this works is that each hidden node basis function has local region of influence
> > and a 1-to-1 correspondence with a previous worst classified training vector.
----SNIP
> > > >
> > > > Then simultaneously train on samples or characteristic exemplars from all 14.
> > >
> > > > If all of the data is not available at once, do it in stages.
> > >
> > > I have all the training and test data, but i dont know how could i do to train 14
> > training vectors and then validate it with just 1 set to check if the neural net is generalizing well.

Train on a subsample of all 14 data sets.

> > > Tying to be clear about wat i'm doing. here is the code:
> > >
> > > ia=linear_train_1(1:5001,4);
> > > w=linear_train_1(1:5001,5);
> > > tq=linear_train_1(1:5001,2);
> > > T1=[198:0.000799840032:202]; % Voltage is between 198V and 202V
> > > iateste1=ia_lin_1(1:5001,4);
> > > wteste1=ia_lin_1(1:5001,5);
> >
> > Seems finely spaced. Do you reallyy need this much data? See below.
> >
> > > P=[T1;ia';w']; % This is the training vector that in this case, trains just 1 set of data.
> > > T=[tq']; I want my neural net to recognize 14 samples of [T1;ia;w']. T1 is fix but
> > 'ia'' and 'w'' varies according to the load equation i'm changing on my motor model. The
> > question is How could i train it to recognize those 14 samples? If i make 'Ia'' and 'w'' a
> > matrix of 14 different currents and speed, this neural net do not allow me to test a simple
> > >vector like is below
> >
> > You need to test matrices not single vectors..
> >
> > > net=feedforwardnet([5 25],'trainbr');
> >
> > Why 2 hidden layers??? Why H =25 ?? Why 'trainbr?
> >trainbr: i have to use bayesien-regulation

You did not answer WHY you think you need to use trainbr !

> > > net.trainParam.goal = 0.005; %error
> >
> > Why?
> defined goal.

But WHY is it the number 0.005 ? Why not 0.01 or 0.0001?

> > > net.trainParam.epochs = 2000;
> >
> > Why?
> maximum training epochs that i want, if not reached the trainparam.goal.
    
How did you determine that the default value is insufficient?
    
> > > net=train(net,P,T);
> >
> > Performance evaluation??
> >
> > [ net tr ] = ...
> > MSEtrn = ?
> > MSEval =?
> > MSEtst = ?
>
> I dont understand what it means.

Always include the structure tr when training. It includes a plethora of training information.
In particular, the MeanSquareErrors for trn, val and tst subsets. To see what I mean just
type the following after the training command:

tr = tr

> > Otherwise, how do you obtain separate tr/val/tst results.
> >
> > > P1=[T1;iateste1';wteste1'];
> > > Y = sim(net,P1);
> > >
> > > As you can see, i'm not an expert on this ... i imagine if you could help me build this
> > process of train and validate. Thanks a lot for your help!
> >
> > This is post No. 8 of this thread and you don't seem to be any further along than you were
> > at the first post. So, let's start again
>
> It's a dificult task to explain. The goal of this thread is (code by code) determine how to train
> different sets of [V;Ia;w] defined above, so that my neural net will recognize those 14 datas.

No. This is not a pattern recognition task. You are not trying to develop a 14-class classifier.

You are trying to develop a regression or curvefitting model that estimates the motor load from 14 samples of voltage, current and motor speed measurements.

> > 1. What is a motor model?
> Simulink-SimPowerSystems
>
> There's an induction motor machine model. Resuming, i'm extracting data from this model
> associated with other procediment that does not matter here..
>
> > 2. What is a motor load?
>
> A motor is a device that converts electrical energy into mechanical energy to act upon a
> mechanical load. The burden placed on the motor due to this mechanical activity is referred to as the motor load.
>
> > 3. What are V, ia, w and tq ?
> V -> Voltage
> ia -> Current on phase 'a'
> w-> motor speed
> tq-> it's the load generated according to the type of burden used to train.

Somehow it bothers me to mix up lower case and capitals when choosing variable names. Also, I like to use variable names that someone not familiar with the work can look at and immediately know what it is without going back to look at the definition. For example, s or ms for motor speed.

Is tq torque?

> > 4.What are the corresponding correlation coefficients?
> I think that does'nt matter too

WRONG, WRONG, WRONG.

Corr coefs tell the complete story in a linear model where the variables are standardized.
Although this is not true for linear models, in addition to I/O plots, it usually is the first
quantitative clue as to (1) whether or not a particular input variable is significant for estimating a particular output variable. (2) what the apparent significance rankings are for the inputs.

Not only do I obtain the correlation coefficient matrix for all variables, I also create the
BACKSLASH linear model. Sometimes I also create a reduced variable linear model
using STEPWISEFIT in a backward search mode.

I have found this info useful in a number of difficult NNET designs.

> > 5. What , exactly, are the differences between the 14 data sets?
> Defined what load is, the difference between the 14 data sets is the type with burden
> i'm using on the motor.
>
> > 6. Have you plotted the output to determine how much sample spacing is
> > needed to adequately characterize it?
> 5000 datas is enough to have values from the transitory state to steady state.

This question is about spacing. For example, could you halve the spacing and use
2500 measurements?, etc.

> > 7. Given that spacing, how much data is needed for that characterization?
> 6.

Is 6 an answer or a typo? If the former, what does it mean?

> > 8. Your first post mentions 10,006 measurements but later you use 5,0001.
> Yes. I've cut unnecessary data.
>
> > Is that for each of the 14 data sets?
> one data set is a value of V ; Ia; W. Only 'V' is fix. Ia nd W varies in each of the
14 data sets. There are 5001 values of Ia and 5001 Values of W as there are 5001
>values of fixed V (voltage)

OK. Just checking.Typically, the terminology "data set" refers to the complete collection
of input and output variables for one or more operating conditions.

> > 9. As I stated before
> > 1. Only 1 hidden layer is necessary
> > 2. If you have 14 scenarios that you want to characterize with one net:
> > a. Take 6 and 7 into consideration and combine samples of all 14 into
> > multiple mixed subsets.
> > b. Since you have a large data set, Train/Validate and Test with a
> > 0.34/0.33/0.33 data split.
> > c. Use one or more data sets, as many defaults as possible, and vary
> > H to find the minimum acceptable value.
> >
> > This should give you a solid start.
> >

Thanks for answering the questions. (Again, why trainbr ?)

Hope this helps.
 
 Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 5 Nov, 2012 20:48:08

Message: 11 of 20

"Carlos Aragon" wrote in message <k742iq$il8$1@newscl01ah.mathworks.com>...
> "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6vg6s$36r$1@newscl01ah.mathworks.com>...
> > "Carlos Aragon" wrote in message <k6ut1o$1mo$1@newscl01ah.mathworks.com>...
> > > "Greg Heath" <heath@alumni.brown.edu> wrote in message <k6rgsg$sp3$1@newscl01ah.mathworks.com>...
> > > > PLEASE, PLEASE DO NOT TOP POST!!!
> > > >
> > > > "Carlos Aragon" wrote in message <k6mhuh$etk$1@newscl01ah.mathworks.com>...
> > > > > Greg, thanks in advance. You're helping a lot!
> > > > >
> > > > > You said:
> > > > >
> > > > > (..)
> > > > >
> > > > > The best is to use a modication of NEWRB that allows the input of an initial
> > > > > > hidden layer. Then
> > > > > >
> > > > > > 1. After training with set1, use those weights as initial weights for training with set2 + set1.
> > > >
> > > > or, if you are lucky
> > > >
> > > > > > 2. After training with set1, use those weights as initial weights for training with set2 and a
> > "characteristic subset" of set1. The drawback is how to define that characteristic.
> > > > > >
> > > > > > The reason this works is that each hidden node basis function has local region of influence
> > and a 1-to-1 correspondence with a previous worst classified training vector.
> > > > >
> > > > > (...)
> > > > >
> > > > > I'm facing problems to perform this action on matlab.
> > > >
> > > > That statement is absolutely useless. I thought you wanted my help.
> > > >
> > > > > Is there any automated way there i can record set1 and then use it to train a set2?
> > > >
> > > > I have no idea what the second part of that statement means.
> > > >
> > > > >How could i do it? Actualy, i want my feedforwardnet to recognize 14 sets of diferent motor loads.
> > > >
> > > > Then simultaneously train on samples or characteristic exemplars from all 14.
> > >
> > > > If all of the data is not available at once, do it in stages.
> > >
> > > I have all the training and test data, but i dont know how could i do to train 14
> > training vectors and then validate it with just 1 set to check if the neural net is generalizing well.
> >
> > Not even close. See below.
> >
> > > Tying to be clear about wat i'm doing. here is the code:
> > >
> > > ia=linear_train_1(1:5001,4);
> > > w=linear_train_1(1:5001,5);
> > > tq=linear_train_1(1:5001,2);
> > > T1=[198:0.000799840032:202]; % Voltage is between 198V and 202V
> > > iateste1=ia_lin_1(1:5001,4);
> > > wteste1=ia_lin_1(1:5001,5);
> >
> > Seems finely spaced. Do you reallyy need this much data? See below.
> >
> > > P=[T1;ia';w']; % This is the training vector that in this case, trains just 1 set of data.
> > > T=[tq']; I want my neural net to recognize 14 samples of [T1;ia;w']. T1 is fix but
> > 'ia'' and 'w'' varies according to the load equation i'm changing on my motor model. The
> > question is How could i train it to recognize those 14 samples? If i make 'Ia'' and 'w'' a
> > matrix of 14 different currents and speed, this neural net do not allow me to test a simple
> > >vector like is below
> >
> > You need to test matrices not single vectors..
> >
> > > net=feedforwardnet([5 25],'trainbr');
> >
> > Why 2 hidden layers??? Why H =25 ?? Why 'trainbr?
> >trainbr: i have to use bayesien-regulation
> > > net.trainParam.goal = 0.005; %error
> >
> > Why?
> defined goal.
> > > net.trainParam.epochs = 2000;
> >
> > Why?
> maximum training epochs that i want, if not reached the trainparam.goal.
> > > net=train(net,P,T);
> >
> > Performance evaluation??
> >
> > [ net tr ] = ...
> > MSEtrn = ?
> > MSEval =?
> > MSEtst = ?
>
> I dont understand what it means.
>
> > Otherwise, how do you obtain separate tr/val/tst results.
> >
> > > P1=[T1;iateste1';wteste1'];
> > > Y = sim(net,P1);
> > >
> > > As you can see, i'm not an expert on this ... i imagine if you could help me build this
> > process of train and validate. Thanks a lot for your help!
> >
> > This is post No. 8 of this thread and you don't seem to be any further along than you were
> > at the first post. So, let's start again
>
> It's a dificult task to explain. The goal of this thread is (code by code) determine how to train different sets of [V;Ia;w] defined above, so that my neural net will recognize those 14 datas.
>
> > 1. What is a motor model?
> Simulink-SimPowerSystems
>
> There's an induction motor machine model. Resuming, i'm extracting data from this model associated with other procediment that does not matter here..
>
> > 2. What is a motor load?
>
> A motor is a device that converts electrical energy into mechanical energy to act upon a mechanical load. The burden placed on the motor due to this mechanical activity is referred to as the motor load.
>
> > 3. What are V, ia, w and tq ?
> V -> Voltage
> ia -> Current on phase 'a'
> w-> motor speed
> tq-> it's the load generated according to the type of burden used to train.
>
> > 4.What are the corresponding correlation coefficients?
> I think that does'nt matter too
>
> > 5. What , exactly, are the differences between the 14 data sets?
> Defined what load is, the difference between the 14 data sets is the type with burden i'm using on the motor.
>
> > 6. Have you plotted the output to determine how much sample spacing is
> > needed to adequately characterize it?
> 5000 datas is enough to have values from the transitory state to steady state.
>
> > 7. Given that spacing, how much data is needed for that characterization?
> 6.
> > 8. Your first post mentions 10,006 measurements but later you use 5,0001.
> Yes. I've cut unnecessary data.
>
> > Is that for each of the 14 data sets?
> one data set is a value of V ; Ia; W. Only 'V' is fix. Ia nd W varies in each of the 14 data sets. There are 5001 values of Ia and 5001 Values of W as there are 5001 values of fixed V (voltage)
>
> > 9. As I stated before
> > 1. Only 1 hidden layer is necessary
> > 2. If you have 14 scenarios that you want to characterize with one net:

Yes. That is exactly what i want.

> > a. Take 6 and 7 into consideration and combine samples of all 14 into
> > multiple mixed subsets.

How do i combine those samples? How do i make this "multiple mixed subsets"?

> > b. Since you have a large data set, Train/Validate and Test with a
> > 0.34/0.33/0.33 data split.
Ok!
> > c. Use one or more data sets, as many defaults as possible, and vary
> > H to find the minimum acceptable value.
> >
> > This should give you a solid start.
> >
> > Hope this helps.
> >
> > Greg

Thanks.

Carlos.

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 17 Nov, 2012 01:54:09

Message: 12 of 20

Greg,

During this week i've been in an exhaust mood to train the neural net. I want to know how could i validate the trained neural network simulating with sim(net,P(test)) considering that the matrix P(test) does not have the same size as the input. is it possible? How?
Every time when i try to simply validate a test input matrix with a differente size i get an error.

Thanks.

Carlos.

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 17 Nov, 2012 10:27:08

Message: 13 of 20

"Carlos Aragon" wrote in message <k86qo1$1vl$1@newscl01ah.mathworks.com>...
> Greg,
>
> During this week i've been in an exhaust mood to train the neural net. I want to know how could i validate the trained neural network simulating with sim(net,P(test))

You are using incorrect terminology.

ASSUMPTION: Training, Validation and Test data can be assumed to be
                       random draws from the same probability distribution.
TRAINING: Used to estimate weight/bias values
VALIDATION(1): Used during training to determine when overtraining an overfit net begins
                          to degenerate the ability to generalize to nontraining data.
VALIDATION(2): Used after training to rank multiple designs
TEST: Used last to estimate performance on unseen data

considering that the matrix P(test) does not have the same size as the input. is it possible? How?
> Every time when i try to simply validate a test input matrix with a differente size i get an error.

If you train a net with node topology I-H-O, the size of any input/target pair must have the form

size(input) = [ I N ]

size(target) = [ O N ]

The resulting output will have the same size as the target.

If you apply the net to unknown data with or without a known target matrix, the input/output pair nust have the form

size(newinput) = [ I Nnew]

size(newoutput) = [ O Nnew]

Hope this helps.

Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 17 Nov, 2012 11:57:08

Message: 14 of 20

Greg,

Next i present you snippets of text from a paper named " Neural Approach for Automatic Identification of Induction Motor Load Torque " from the author Alessandro Goedtel. All i'm doing is trying to redo the procedure. I think you can help me build the process according to it.
Remember: Inputs: Voltage , Current and Speed
                Output: Torque
                Training Curves (extracted from simulink): 13
                Testing Curves for validation: 13
Note: He used 100 pairs of input-output .. i'm using 5000 pairs.

>>(...)Each curve set, simulating a specific load from 5% to 250% of nominal torque, is >>composed of a vector constituted of 100 input-output pairs, which represent the >>transient and steady behavior of the motor for one specific load and a voltage range.(...)

>>(...) During the training process, the input-output pairs representing the process behavior >>are sequentially presented to the network(...)

What can i understand from"sequentially present" the pairs input-output? Could you give an example on how to do it?

>>(...)For the case of a quadratic load and a specific voltage
>>range (198V,...,242V) 26 simulations (curves) were generated, which simulate the
>>motor's behavior from 5% to 250% of nominal torque, where 13 were used for the >>training process and 13 for the testing process. Each training set, composed of 13 >>curves, is constituted by 100 input-output pairs, which are sequentially grouped to >>produce the training matrix. Each voltage range and each load type has a specific >>neural network.(...)
>>After the training process, the network is able to estimate
>>load torque curve from sequential values of speed, current and
>>voltage. In this case, the testing process used to validate the
>>proposed approach consists of using other operating
>>configurations that were absent during the training process.(...)

Ok. Once i have done for one type of load, i can do for all the other types.
The problem: i'm still wondering how he built the training process for those 13 curves and even how he could validate for other 13 curves. All the tips given from above is that he used "other operating configurations that were absent during the training process".
What kind of "configuration" could be it? (Show me an example, please)
I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.
And sorry for my bad english anyway...

Thanks!

Carlos

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 17 Nov, 2012 17:14:14

Message: 15 of 20

"Carlos Aragon" wrote in message <k87u2k$ng9$1@newscl01ah.mathworks.com>...
> Greg,
>
> Next i present you snippets of text from a paper named " Neural Approach for Automatic Identification of Induction Motor Load Torque " from the author Alessandro Goedtel. All i'm doing is trying to redo the procedure. I think you can help me build the process according to it.
> Remember: Inputs: Voltage , Current and Speed
> Output: Torque
> Training Curves (extracted from simulink): 13
> Testing Curves for validation: 13
> Note: He used 100 pairs of input-output .. i'm using 5000 pairs.
>
> >>(...)Each curve set, simulating a specific load from 5% to 250% of nominal torque, is >>composed of a vector constituted of 100 input-output pairs, which represent the >>transient and steady behavior of the motor for one specific load and a voltage range.(...)
>
> >>(...) During the training process, the input-output pairs representing the process behavior >>are sequentially presented to the network(...)
>
> What can i understand from"sequentially present" the pairs input-output? Could you give an example on how to do it?
>
> >>(...)For the case of a quadratic load and a specific voltage
> >>range (198V,...,242V) 26 simulations (curves) were generated, which simulate the
> >>motor's behavior from 5% to 250% of nominal torque, where 13 were used for the >>training process and 13 for the testing process. Each training set, composed of 13 >>curves, is constituted by 100 input-output pairs, which are sequentially grouped to >>produce the training matrix. Each voltage range and each load type has a specific >>neural network.(...)
> >>After the training process, the network is able to estimate
> >>load torque curve from sequential values of speed, current and
> >>voltage. In this case, the testing process used to validate the
> >>proposed approach consists of using other operating
> >>configurations that were absent during the training process.(...)
>
> Ok. Once i have done for one type of load, i can do for all the other types.
> The problem: i'm still wondering how he built the training process for those 13 curves and even how he could validate for other 13 curves. All the tips given from above is that he used "other operating configurations that were absent during the training process".
> What kind of "configuration" could be it? (Show me an example, please)
> I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.
> And sorry for my bad english anyway...

Why haven't you contacted the authors of the paper?

Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 17 Nov, 2012 21:09:17

Message: 16 of 20

"Greg Heath" <heath@alumni.brown.edu> wrote in message <k88gl6$mn5$1@newscl01ah.mathworks.com>...
> "Carlos Aragon" wrote in message <k87u2k$ng9$1@newscl01ah.mathworks.com>...
> > Greg,
> >
> > Next i present you snippets of text from a paper named " Neural Approach for Automatic Identification of Induction Motor Load Torque " from the author Alessandro Goedtel. All i'm doing is trying to redo the procedure. I think you can help me build the process according to it.
> > Remember: Inputs: Voltage , Current and Speed
> > Output: Torque
> > Training Curves (extracted from simulink): 13
> > Testing Curves for validation: 13
> > Note: He used 100 pairs of input-output .. i'm using 5000 pairs.
> >
> > >>(...)Each curve set, simulating a specific load from 5% to 250% of nominal torque, is >>composed of a vector constituted of 100 input-output pairs, which represent the >>transient and steady behavior of the motor for one specific load and a voltage range.(...)
> >
> > >>(...) During the training process, the input-output pairs representing the process behavior >>are sequentially presented to the network(...)
> >
> > What can i understand from"sequentially present" the pairs input-output? Could you give an example on how to do it?
> >
> > >>(...)For the case of a quadratic load and a specific voltage
> > >>range (198V,...,242V) 26 simulations (curves) were generated, which simulate the
> > >>motor's behavior from 5% to 250% of nominal torque, where 13 were used for the >>training process and 13 for the testing process. Each training set, composed of 13 >>curves, is constituted by 100 input-output pairs, which are sequentially grouped to >>produce the training matrix. Each voltage range and each load type has a specific >>neural network.(...)
> > >>After the training process, the network is able to estimate
> > >>load torque curve from sequential values of speed, current and
> > >>voltage. In this case, the testing process used to validate the
> > >>proposed approach consists of using other operating
> > >>configurations that were absent during the training process.(...)
> >
> > Ok. Once i have done for one type of load, i can do for all the other types.
> > The problem: i'm still wondering how he built the training process for those 13 curves and even how he could validate for other 13 curves. All the tips given from above is that he used "other operating configurations that were absent during the training process".
> > What kind of "configuration" could be it? (Show me an example, please)
> > I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.
> > And sorry for my bad english anyway...
>
> Why haven't you contacted the authors of the paper?
Belieave me, i really have tried.
> Greg

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 18 Nov, 2012 02:47:16

Message: 17 of 20

"Carlos Aragon" wrote in message <k87u2k$ng9$1@newscl01ah.mathworks.com>...
> Greg,
>
> Next i present you snippets of text from a paper named " Neural Approach for Automatic Identification of Induction Motor Load Torque " from the author Alessandro Goedtel. All i'm doing is trying to redo the procedure. I think you can help me build the process according to it.
> Remember: Inputs: Voltage , Current and Speed
> Output: Torque
> Training Curves (extracted from simulink): 13
> Testing Curves for validation: 13
> Note: He used 100 pairs of input-output .. i'm using 5000 pairs.

I don't know what that means. You have 3 input variables and 1 output variable.
An input-output pair consists of a 3-D input and the corresponding 1-D output?
100 such pairs make up a curve? But what constitutes "a" curve when you have 3 input variables? Are two inputs held constant and the curve consists of the output vs the one input that is varying? Or do you have 4 time series?...

> >>(...)Each curve set, simulating a specific load from 5% to 250% of nominal torque, is >>composed of a vector constituted of 100 input-output pairs, which represent the >>transient and steady behavior of the motor for one specific load and a voltage range.(...)

Not clear.
 
> >>(...) During the training process, the input-output pairs representing the process behavior >>are sequentially presented to the network(...)

Is this a time series?
 
> What can i understand from"sequentially present" the pairs input-output? Could you give an example on how to do it?

I don't know. It is probably a time series. The only other possibility is sequential rather than batch learning.

> >>(...)For the case of a quadratic load and a specific voltage
> >>range (198V,...,242V) 26 simulations (curves) were generated, which simulate the
> >>motor's behavior from 5% to 250% of nominal torque, where 13 were used for the >>training process and 13 for the testing process. Each training set, composed of 13 >>curves, is constituted by 100 input-output pairs, which are sequentially grouped to >>produce the training matrix. Each voltage range and each load type has a specific >>neural network.(...)

So there are 13 networks? One for each "curve"? Again, what is plotted vs what for each curve?

> >>After the training process, the network is able to estimate
> >>load torque curve from sequential values of speed, current and
> >>voltage. In this case, the testing process used to validate the
> >>proposed approach consists of using other operating
> >>configurations that were absent during the training process.(...)
>
> Ok. Once i have done for one type of load, i can do for all the other types.
> The problem: i'm still wondering how he built the training process for those 13 curves and even how he could validate for other 13 curves. All the tips given from above is that he used "other operating configurations that were absent during the training process".
> What kind of "configuration" could be it? (Show me an example, please)

My guess he had data for 26 curves, ordered them somehow and used every other one for
training.

> I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.

I don't know. It looks like he designed 13 nets. Maybe inputs are presented to all 13 and the target that is closest to the output identifies the correspoding load condition.

That is my best guess.

Greg

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 18 Nov, 2012 22:01:13

Message: 18 of 20

"Greg Heath" <heath@alumni.brown.edu> wrote in message <k89i7j$9un$1@newscl01ah.mathworks.com>...
> "Carlos Aragon" wrote in message <k87u2k$ng9$1@newscl01ah.mathworks.com>...
> > Greg,
> >
> > Next i present you snippets of text from a paper named " Neural Approach for Automatic Identification of Induction Motor Load Torque " from the author Alessandro Goedtel. All i'm doing is trying to redo the procedure. I think you can help me build the process according to it.
> > Remember: Inputs: Voltage , Current and Speed
> > Output: Torque
> > Training Curves (extracted from simulink): 13
> > Testing Curves for validation: 13
> > Note: He used 100 pairs of input-output .. i'm using 5000 pairs.
>
> I don't know what that means. You have 3 input variables and 1 output variable.
> An input-output pair consists of a 3-D input and the corresponding 1-D output?
> 100 such pairs make up a curve? But what constitutes "a" curve when you have 3 input variables? Are two inputs held constant and the curve consists of the output vs the one input that is varying? Or do you have 4 time series?...
>
> > >>(...)Each curve set, simulating a specific load from 5% to 250% of nominal torque, is >>composed of a vector constituted of 100 input-output pairs, which represent the >>transient and steady behavior of the motor for one specific load and a voltage range.(...)
>
> Not clear.
>
> > >>(...) During the training process, the input-output pairs representing the process behavior >>are sequentially presented to the network(...)
>
> Is this a time series?
>
> > What can i understand from"sequentially present" the pairs input-output? Could you give an example on how to do it?
> I don't know. It is probably a time series. The only other possibility is sequential rather than batch learning.

Each set of [Voltage;Current;Speed] gives me a valor of target: 'torque'. In the case i'm using 13 different sets of [Voltage;Current;Speed] and i want, my unique feedforwardnet, to train for 13 different targets corresponding to each of the training inputs, using either batch or senquential training, how can i build it into matlab?

> > >>(...)For the case of a quadratic load and a specific voltage
> > >>range (198V,...,242V) 26 simulations (curves) were generated, which simulate the
> > >>motor's behavior from 5% to 250% of nominal torque, where 13 were used for the >>training process and 13 for the testing process. Each training set, composed of 13 >>curves, is constituted by 100 input-output pairs, which are sequentially grouped to >>produce the training matrix. Each voltage range and each load type has a specific >>neural network.(...)
>
> So there are 13 networks? One for each "curve"? Again, what is plotted vs what for each curve?
Just one network for each type of load.
Torque is the curve. The parameters Voltage, current and speed in time, defines the torque.

> > >>After the training process, the network is able to estimate
> > >>load torque curve from sequential values of speed, current and
> > >>voltage. In this case, the testing process used to validate the
> > >>proposed approach consists of using other operating
> > >>configurations that were absent during the training process.(...)
> >
> > Ok. Once i have done for one type of load, i can do for all the other types.
> > The problem: i'm still wondering how he built the training process for those 13 curves and even how he could validate for other 13 curves. All the tips given from above is that he used "other operating configurations that were absent during the training process".
> > What kind of "configuration" could be it? (Show me an example, please)
>
> My guess he had data for 26 curves, ordered them somehow and used every other one for
> training.
Yes, that is the understanding. But How could he order those 13 for training and the other 13 for test? That is the point.
> > I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.
>
> I don't know. It looks like he designed 13 nets. Maybe inputs are presented to all 13 and the target that is closest to the output identifies the correspoding load condition.
>
> That is my best guess.
>
> Greg

Thansk.

Carlos

Subject: Training multiple data for a single feedforwardnet

From: Greg Heath

Date: 19 Nov, 2012 02:11:11

Message: 19 of 20

"Carlos Aragon" wrote in message <k8blr8$f8p$1@newscl01ah.mathworks.com>...
> "Greg Heath" <heath@alumni.brown.edu> wrote in message <k89i7j$9un$1@newscl01ah.mathworks.com>...
> > "Carlos Aragon" wrote in message <k87u2k$ng9$1@newscl01ah.mathworks.com>...
======SNIP
> > My guess he had data for 26 curves, ordered them somehow and used every other one for
> > training.
> Yes, that is the understanding. But How could he order those 13 for training and the other 13 for test? That is the point.

If I were him I would overlay the 13 torque curves and try to pick a representative sample.

Greg
> > > I'm concerned on how to build all the training and testing process into the neural net. Hope it's more clear for you to understand what i'm trying to do. If you can show me examples for your exaplanation it would be easier for me to understand.
> >
> > I don't know. It looks like he designed 13 nets. Maybe inputs are presented to all 13 and the target that is closest to the output identifies the correspoding load condition.
> >
> > That is my best guess.
> >
> > Greg
>
> Thansk.
>
> Carlos

Subject: Training multiple data for a single feedforwardnet

From: Carlos Aragon

Date: 1 Dec, 2012 23:14:08

Message: 20 of 20

I think i'm almost there ...

How do i record the best training values? Bias, weights etc.

Thanks.

Tags for this Thread

What are tags?

A tag is like a keyword or category label associated with each thread. Tags make it easier for you to find threads of interest.

Anyone can tag a thread. Tags are public and visible to everyone.

Contact us