Simplefit_dataset  has 2 columns [x,t] normally 'x' is considered as input and 't' as target.

feedforwardnet(10)  returns a network of 10 hidden layers + 1 output.

Then you train based on inputs and target output you have provided in (x,t) with the net configurations such as hidden size and some functions .

view etc  etc are for the visualizing it .

perf is for graphs ...

 if you want to change the input then change 'x' and give the 't' on the required output . 

Look at the net e.g: net.< fields > , this is like a structure object , one of the thing is function used (Check help) it will list the configurations of the net that you can change. 

Thanks @ Aparna Sathya Murthy

• My Input is  [311 x 22]  ---That is, I have 311 rows and 22 columns.
• My output is  [311 x 1] --- That is, I have 311 rows and 1 column as Target Output.

The network was trained using the code provided by MATLAB for Elman, Feedforward, and Layer Recurrent.

The network was trained and worked fine.

MY QUESTION:

I have new inputs that I want to predict the output based on the trained network.

The new inputs are not part of the trained network.

I just want to predict the outcome from the new set of inputs.

if 'net' is your trained network and 'z' your new input, then the predictions 'pred' are given by:

pred = net(z)

analogously to y = net(x) in your example above, which gives you the net's predictions for the input it was trained on. Note, that the network must be called with it's variable name from the train() line, e.g.

net = feedforwardnet(10);

myNewNet = train(net,x,t)

pred = myNewNet(z)

 Thanks @ Uwe.

This was my code:

clear

x = inputs 311 x 21; % These are imported from excel

t = outputs 311 x 1;  %  Just wanted to give you the idea of the input and output

x = (dinputs)';

t = (douts)';

net = feedforwardnet(10);

net = train(net,x,t);

view(net)

y = net(x);

perf = perform(net,y,t)

-------------------Network was trained with the above code-----------------------------------

Now:

I will like to see the output for New Inputs:

I have this new input [ 1 x 21] ; that is,1 row , 22 columns.

New Input 

[1, 2, 3 ,4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22]

Question:

How would I use the set of new inputs above to get an OUTPUT from the trained network?

Thanks

Thanks to you all.

Now solved. 

To predict the output for new set of inputs after the network has been trained,

The following command was helpful:

drwiy = net ([[a;b;c]]]

Where:

drwiy == variable name to contain the result.

net= the trained network.

a= value of the 1st column,

b = value of the second column and so on.

THANK you ALL.

can i use feed forward for unsupervised

Abida Khanum I guess this is not appropriate. The most common unsupervised learning method is cluster analysis, which is used for exploratory data analysis to find hidden patterns or grouping in data.

Aldrwiy Hayhar if we want to predict for a new input many times and we dont know how many of them are exactly then how can we predict ?

for example I am using this model in optimization so every time i need to predict output for a new input if i used the above example of

y = net(x_new)

it takes alot of time and train the whole data every time.

is there any possibilty that we can just used the trained model to predict many new outputs?

Rasheed Omobolaji Alabi I am forecasting stock data with one variable.

so to forecast, what is my a & b ? i want to forecast 200 values.

drwiy = net ([[a;b;c]]]

Where:

drwiy == variable name to contain the result.

net= the trained network.

a= value of the 1st column,

b = value of the second column and so on.