how to use deep learning with multivariable?

Hello everyone,

I've been searching for how to use more than one independent variable in deep learning but I couldn't find it yet. Also I'm not sure if it's possible or not. Am i missing something important about deep learning theory? Hope i could explain my problem.

Thanks in advance!!

Answer 1

[Mahesh Taparia]

Hi

It is not clear what kind of data you are having and what you want to do with that.Explain it. Also confirm whether you are having a labeled dataset or unlabeled dataset? The algorithm is completely dataset and application dependent. It might be the case that, it can be solved by some basic machine learning algorithms.

[Meltem Tekinay]

Hello, I'm working with stock prices. So, my dataset is time series and I've made a net with deep learning toolbox in order to predict them. It's working very well, or just seems like that. Now i'm wondering if i could add more independent variables (more datasets as XTrain, XTest, XValid) to my net. Then i'll do comparison with these 2 results.

Thanks a lot!!

[Mahesh Taparia]

Hi

You can save your trained weights and model performance on test data for comparison with different model.

If you want to increase the training dataset, you need to retrain the model/ used the previous weights and start training again. In this case you need to add the data in the training dataset folder/file.

Similarly for the test/validation data, add them into respective source file/folder and then check the performance.

[Meltem Tekinay]

Hello,

Thank you for your response. I want to be sure about something.

>>net = trainNetwork(XTrain,YTrain,layers,options); I use this code in order to train my datas. But I want it with not just one XTrain, I want a model with XTrain and XTrain2, XTrain3, XTrain4 and XTrain5 to just YTrain. So, can I do after that

>>net = trainNetwork(XTrain,YTrain,layers,options);

then not clearing workspace and not deleting "net", apply

>>net = trainNetwork(XTrain1,YTrain,layers,options); and for all other XTrain datas?

[Mahesh Taparia]

I think you want to train different models in a single code with different set of input. In this case, keep the different name of net variable, i.e.

net1 = trainNetwork(XTrain1,YTrain,layers,options); 
net2 = trainNetwork(XTrain2,YTrain,layers,options); 

[Meltem Tekinay]

I want one model that has relation with all XTrain and YTrain datas. As I searched, trainNetwork function is working with same weights and if I use same network, when I add XTrain1 then training is developing. Do I think right?

[Mahesh Taparia]

Hi Club the complete dataset into 1 variable Xtrain instead of separate variables (as you mentioned Xtrain1/ Xtrain2) and then train the model.