How can I train data separated into subsets

[chamrik]

Hi all,

Maybe i shud put it this way. Suppose I want to perform training on part of the data and leave the rest for testing. the training set is separated into subsets whereby I train one subset first, and use its weights and biases as weights to the next subset and so on. Then how to I write that loop taking into consideration that after I present every sample for training I will repeat the same process. Under normal circumstances we do this:

net = newff(mimax(x),[2,2],{'logsig','purelin'},'traingd');
net = train(net,x,t);

But now I want to perform something like this:

net = train(net,xtrain1,ttrain1);
net = train(net,xtrain2,ttrain2);

Maybe at this stage I can simulate and then if need be train again, go back to the above step.

but how do I present this in a form of a loop?

Please help!

 

[Aaron Bex]

You could try something like this:

for i = 1:2
xtrain = x(:,i);
ttrain = t(:,i);
net = train(net,xtrain,ttrain);
end

 

[ravenprp]

Hi there,

Thank you for reaching out with your question. To train data separated into subsets, you can use a for loop to iterate through each subset and train the network accordingly. Here is an example of how you can structure the loop:

net = newff(mimax(x),[2,2],{'logsig','purelin'},'traingd'); % initialize the network

for i = 1:num_subsets % replace num_subsets with the number of subsets you have
% train the network on the current subset
net = train(net, xtrain{i}, ttrain{i}); % replace xtrain{i} and ttrain{i} with the data for the current subset

% use the trained network to make predictions on the current subset's test data
ytest{i} = sim(net, xtest{i}); % replace xtest{i} with the test data for the current subset
end

In this example, each subset is represented by the index i. The loop will run for the specified number of subsets, and within each iteration, the network will be trained on the current subset and then used to make predictions on the test data for that subset. You can then use the predicted values to evaluate the performance of the network on each subset.

I hope this helps. Let me know if you have any further questions. Best of luck with your training!