Changing weight for each Kernel

[always...@gmail.com]

Dear Sir

I am trying to combine autoencoder with CNN.

I used weights those were learned from autoencoder and exchanged those weight with the weight, which were automatically set at net.layers(k).weight(n).

After i input all the "new" weight and run the CNN, following error came out

"Verifying backProp.. Error using assert

The condition input argument must be a scalar logical.

Error in verifyBackProp (ling 99)

                                      assert(0,'How come??, problem in layer %d,k);

Error in Train (line 39)

              verifyBackProp(net);

Error in untitled55 (line 123)

NET=Train(MNIST3d,net,100);

I tried to debug the error but i could not find the solution (do i have to further change other setting or input? when i change the weight of kernel?)

Do you know what seem to make this error? 

Thank you again 

Best regard

Hang-Rai Kim

[Hagay Garty]

You can disable the backprop verification by adding

net.runInfoParam.verifyBP       = 0;

to your config file.

Anyway, you should not get these errors if all is OK. The VerifyBP function is some kind of sanity to make sure the derivatives are calculated well. If it fails this means probably something is wrong.

Can you describe in more details what exactly are you doing..