Recurrent Neural Networks(RNN) are a type of Neural Network where the output from the previous step is fed as input to the current step. In the case of Convolution Neural Networks (CNN), the output from the softmax layer in the context of image classification is entirely independent of the previous input image.