CNN architecture pattern - LeNet

LeNet is a popular architectural pattern for implementing CNN. In this chapter, we shall learn to build CNN model based on LeNet pattern by creating the layers in the following sequence:

1. The input layer
2. The convolutional layer 1 that produces a set of feature maps, with ReLU activation
3. The pooling layer 1 that produces a set of statistically aggregated feature maps
4. The convolutional layer 2 that produces a set of feature maps, with ReLU activation
5. The pooling layer 2 that produces a set of statistically aggregated feature maps
6. The fully connected layer that flattens the feature maps, with ReLU activation
7. The output layer that produces the output by applying simple linear activation

Note

LeNet family of models were introduced by Yann LeCun and his fellow researchers. More details on the LeNet family of models can be found at the following link: http://yann.lecun.com/exdb/publis/pdf/lecun-01a.pdf.

 

Yann LeCun maintains a list of the LeNet family of models at the following...