Testing real data in NN

[allinjoe]

I trained and validated the datas in NN but now i need to test the real data in NN. Kindly help.

[wangruinan]

how can I help you? please give me an email in detail to clarify your question.

[allinjoe]

Dear Sir, Thanks for your Kind Consideration on my issue. I had 100 samples. I trained my NN.m with 80 samples and validated using the remaining 20 samples. Now I need to give real data other than those 100 samples used for training and validation. I will give only the input values of samples and the system needs to find the output values based on the training provided. I felt that it was not possible using your toolbox. Help me in resolving this issue. Thank You.

…

On Mon, Jun 26, 2017 at 1:17 PM, wangruinan ***@***.***> wrote: how can I help you? please give me an email in detail to clarify your question. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#175 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AcR3ZhM-ZIHYcgZpt1ZbTfnXAAoYOjrYks5sH2IJgaJpZM4ODRnS> .

-- Regards, Mr.D.Allin Joe Assistant Professor - I, Department of ECE, Kumaraguru College of Technology, Coimbatore - 641049. <[email protected]>

[wajihullahbaig]

@allinjoe If you want to test it on real data. You still need training and testing.
Say you have 800 samples. You give 600 for training and 200 for testing nowl
a) Once you have trained the network, you need to save the neural network and then load it to test it on anther sample set.
b) Say this other sample set is of 500 images that you want to test.
c) Load the neural network (trained on 600 images), then just test push your new test samples (500 images) and see visually if it does a correct prediction. If you have labeled the data, you can make a comparison against the labeled data and the NN predictions.

Hope this helps.

[allinjoe]

Sir,
Where can i find the output values of validation data before mse calculation? kindly help

[lcayer]

Thank you for your answer！ 在2017年06月29 12时47分, "wajihullahbaig"<[email protected]>写道: @allinjoe If you want to test it on real data. You still need training and testing. Say you have 800 samples. You give 600 for training and 200 for testing nowl a) Once you have trained the network, you need to save the neural network and then load it to rest it on anther sample set. b) Say this other sample set is of 500 images that you want to test. c) Load the neural network (trained on 600 images), then just test push your new test samples (500 images) and see visually if it does a correct prediction. If you have labeled the data, you can make a comparison against the labeled data and the NN predictions. Hope this helps. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[allinjoe]

TRAINING DATA.xlsx

Kindly reply this issue with the attached excel file not for images.

[wajihullahbaig]

@allinjoe I dont understand the excel file.
No one is going to do it for you except yourself. Now following the path to tests folder
https://github.com/rasmusbergpalm/DeepLearnToolbox/blob/master/tests/test_example_NN.m

And see the line#20 - [er, bad] = nntest(nn, test_x, test_y);
Follow the function and see how the errors are calculated. This is where labeled data is compared against predictions.