The SageMaker image classification algorithm is an implementation of residual networks (ResNets). Before we delve into the details of the algorithm, let's briefly understand convolutional neural networks (CNN) and ResNet and how they learn patterns from images.
Like any other neural network, CNNs are made up of input, hidden, and output layers. These networks have learnable parameters called weights and biases. These weights and biases can be adjusted through an appropriate optimizer, such as Stochastic Gradient Descent (SGD), with backpropagation. However, the difference between any feedforward artificial neural network and CNNs is that the hidden layers in CNNs are convolutional layers. Each convolutional layer consists of one or more filters. The job of these filters is to recognize patterns in input images.
These...
