Member-only story
Going Deeper with CNN: Understanding Layers, Nodes, Kernels and Backpropagation
If you’ve come across this article while trying to understand Convolutional Neural Networks (CNNs), it would be beneficial for you to begin by reading my previous article:
Now, let’s delve deeper into CNN by gaining a better understanding of layers and how they are interconnected and communicate. In general, there are three primary types of layers: (1) the input layer, (2) hidden layers, and (3) the output layer.
→ The Input Layer: serves as the entry point for the input data and prepares it for subsequent layers. This initial step may involve altering the data format and size to suit the network’s requirements. It’s important to note that the input layer is typically regarded as a data preprocessing layer and does not possess any learnable parameters.
→ Hidden Layers: refer to all the layers positioned between the input layer and the output layer. These layers are termed “hidden” because we cannot directly observe their outputs. It is within these hidden layers that the learnable parameters come into play and capture data representations. Think of the hidden layers as the layers responsible for extracting and transforming features from the data. A hidden layer receives information from the preceding layer and transmits information to the subsequent layer.
→ Output Layer: serves as the final layer in the network and produces a result or prediction based on the learnings acquired from the preceding layers. The output layer encompasses both the activation function, which introduces non-linearity to the output, and the loss function, which quantifies the discrepancy between the predicted output and the actual output.
You can learn more about Loss Functions here:
A convolution layer is a hidden layer!
