CS 451 Quiz 33
Convolutional nets
* Required
Email address
*
Your email
In a convolutional layer, the weights in the convolution filters play a similar role as the weight matrix W in a fullyconnected layer
*
True
False
Given a 6x6x3 input image, what is the shape of the output of a convolutional layer that applies 5 filters of shape 3x3x3?
*
4 x 4 x 3
4 x 4 x 5
4 x 4 x 15
Unlike in a fullyconnected layer, in a convolutional layer we don't have any bias parameters
*
True
False
What types of layers does a convolutional net typically contain? Check all that apply
*
Fully connected
Convolution
Scrambling
Pooling
Required
Given the matrix [1 2 3; 4 5 6; 7 8 9], what is the output of a maxpooling layer with filter size f=2 and stride s=1?
*
[5 5; 8 8]
[3 4; 6 7]
[5 6; 8 9]
[1 3; 7 9]
A pooling layer has no parameters that need to be learned
*
True
False
If you start from the input layer and go from layer to layer deeper into a CNN, what typically happens?
*
The image size increases and the number of channels increases
The image size increases and the number of channels decreases
The image size decreases and the number of channels increases
The image size decreases and the number of channels decreases
What is a typical sequence of layers?
*
conv, conv, conv, pool, pool, pool, fc, fc, fc
conv, pool, fc, conv, pool, fc, conv, pool, fc
conv, pool, conv, pool, conv, pool, fc, fc, fc
Which layers have the most parameters, which the least? Order from most (left) to least (right)
*
conv, pool, fc
fc, conv, pool
conv, fc, pool
fc, pool, conv
Which of the following does NOT explain why convolutions are a good idea?
*
Translational invariance
Rotational invariance
Parameter sharing
Sparsity of connections
Submit
This content is neither created nor endorsed by Google.
Report Abuse

Terms of Service

Additional Terms
Forms