Summary
In this chapter, we learned about CNNs. Specifically, we covered convolution layers and pooling layers (the basic modules that constitute CNNs) in great detail in order to understand them at the implementation level. CNNs are mostly used when looking at data regarding images. Please ensure that you understand the content of this chapter before moving on.
In this chapter, we learned about the following:
- In a CNN, convolution, and pooling layers are added to the previous network, which consists of fully connected layers.
- You can use
im2col(a function for expanding images into arrays) to implement convolution and pooling layers simply and efficiently. - Visualizing a CNN enables you to see how advanced information is extracted as the layer becomes deeper.
- Typical CNNs include LeNet and AlexNet.
- Big data and GPUs contribute significantly to the development of deep learning.
