Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
400 views22 pages

Varience of Basic Convolution Function

The document discusses various variants of convolution functions, emphasizing the limitations of using a single kernel for feature extraction in neural networks. It explains the importance of zero-padding in maintaining the spatial dimensions of the output and introduces concepts such as strided convolution, unshared convolution, tiled convolution, and transposed convolution. These techniques enhance the network's ability to extract multiple features and control output size effectively.

Uploaded by

devanand272003
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
400 views22 pages

Varience of Basic Convolution Function

The document discusses various variants of convolution functions, emphasizing the limitations of using a single kernel for feature extraction in neural networks. It explains the importance of zero-padding in maintaining the spatial dimensions of the output and introduces concepts such as strided convolution, unshared convolution, tiled convolution, and transposed convolution. These techniques enhance the network's ability to extract multiple features and control output size effectively.

Uploaded by

devanand272003
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 22

Variants of convolution functions

Mr. Sivadasan E T
Associate Professor
Vidya Academy of Science and Technology, Thrissur
Variants of the Basic Convolution Function

• Convolution with a single kernel can only


extract one kind of feature.

• Usually we want each layer of our


network to extract many kinds of features,
at many locations.
Variants of the Basic Convolution Function
When working with images, we usually
think of the input and output of the
convolution as being 3-D tensors.

With one index into the different channels


and two indices into the spatial coordinates
of each channel.
Variants of the Basic Convolution Function

Software implementations usually work in


batch mode, so they will actually use 4-D
tensors, with the fourth axis indexing
different examples in the batch.
Variants of the Basic Convolution Function
Strided convolution
Strided convolution
Zero Padding

One essential feature of any convolution


network implementation is the ability to
implicitly zero-pad the input V in order to
make it wider.
Zero Padding
One essential feature of any convolution
network implementation is the ability to
implicitly zero-pad the input V in order to
make it wider.

Without this feature, the width of the


representation shrinks by one pixel less than
the kernel width at each layer.
Zero Padding

Zero padding the input allows us to control


the kernel width and the size of the output
independently.
Zero Padding

Without zero padding, we are forced to choose


between shrinking the spatial extent of the
network rapidly and using small kernels - both
scenarios that significantly limit the expressive
power of the network.
Three special cases of the zero-padding setting
1. No zero-padding - Valid convolution.

2. Enough zero-padding is added to keep the size of


the output equal to the size of the input - same
convolution.

3. Enough zeoes are added for every pixel to be visited


k times in each direction, resulting in an output image
of width (m + k -1) - full convolution (m = Image
width)
Zero Padding
Unshared Convolution
Unshared Convolution
Unshared Convolution
Unshared Convolution
Tiled Convolution
Transposed Convolution
• Transposed convolution as the opposite of the
convolution.
• This operation adds all the neighboring numbers in
the input layer together, weighted by a convolution
matrix (kernel). For example, in the image below,
the output value 55 is calculated by the element-
wise multiplication between the 3x3 part of the
input layer and the 3x3 kernel, and sum all results
together:
Transposed Convolution
Thank You!

You might also like