Understanding Keras layers

Keras layers are the fundamental building blocks of Keras models. Each layer receives data as input, does a specific task, and returns an output.

Keras includes a wide range of built-in layers:

• Core layers: Dense, Activation, Flatten, Input, Reshape, Permute, RepeatVector, SpatialDropOut, and many more.
• Convolutional layers for Convolutional Neural Networks: Conv1D, Conv2D, SeparableConv1D, Conv3D, Cropping2D, and many more.
• Pooling layers that perform a downsampling operation to reduce feature maps: MaxPooling1D, AveragePooling2D, and GlobalAveragePooling3D.
• Recurrent layers for recurrent neural networks to process recurrent or sequence data: RNN, SimpleRNN, GRU, LSTM, ConvLSTM2D, etc.
• The embedding layer, only used as the first layer in a model and turns positive integers into dense vectors of fixed size.
• Merge layers: Add, Subtract, Multiply, Average, Maximum, Minimum, and many more.
• Advanced activation...