plz help in this question.
Convolutional Neural QUIZ
Its appreciable that you were very close to answer. You just missed one thing that there are three channels i.e. image is in RGB. So total number of parameters are 3003003*100+ 100
Hope this helped you 
Regarding 2D:
As if there are 3 examples with 4 features,
then we have Weight matrix of shape(no_of_features,no_of_neurons in layer)
and Z matrix of shape(no_of_examples,no_of_neurons),
total parameters in hidden layer=no_of_examplesno_of_featuresno_of_neurons+no_of_neurons(for bias term).
I am unable to relate this with 3D.
Is no_of_examples=1,no_of_features=300300300
if so then how will we write Z matrix means
Is it possible to multiply A[m,n,k] by B[p,q,r]?
Regarding 2D:
Number of hidden parameters is always independent of number of examples.
First of try to understand that by number of parameters we are referring to count of weights and bias terms.
So number of parameters in this case will be = (no_of_features * no_of_neurons)+no_of_neurons(for bias term).
Regarding 3D:
The formula is same as before since this is also a fully connected network.
But in 3D when you input any image with RGB channel, say image is of size 300 * 300, than you can say that total number of features will be 300 * 300 * 3, because there are 3 2-D Matrices of size 300*300 (for each channel, R,G,B).
So total number of parameters are (300 * 300 * 3) * (100) + (100)
got it …
But Z(value of function before applying Activation) is W*X.t +b.
If X is of shape(300,300,3),can we multiply W with X.t ?
Will we first reshape X to ( , )?
For the given question we first need to flatten x , so that’s it final shape becomes, (m, 270000) where m is number of examples, and w is of shape (100,270000) and bias is of shape(100,). So now you can multiply W with x.t directly and add bias to it.
Thank you…
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.