What is the difference between the internal network structures of CNN (Convolutional Neural Network), RNN (Circular Neural Network) and DNN (Deep Neural Network)?

1, DNN: There is a problem-it is impossible to model the change of time series. However, the time series of samples is very important for natural language processing, speech recognition, handwriting recognition and other applications. In order to meet this demand, another neural network structure-RNN came into being.

2.CNN: The signal of each layer of neurons can only propagate to the next layer, and the processing of samples at each moment is independent, so it is also called feedforward neural network.

3.RNN: The output of neurons can directly affect themselves in the next timestamp, that is, the input of layer I neurons at time m, including their own output at time (m- 1) in addition to the output of layer I neurons at that time!

introduce

Neural network technology originated in 1950s and 1960s, when it was called perceptron. It had an input layer, an output layer and an implicit layer. The input feature vector reaches the output layer through hidden layer transformation, and the classification results are obtained at the output layer. The initiator of the early perceptron was Rosenblat.

In practical application, the so-called deep neural network DNN often combines many known structures, including convolution layer or LSTM unit.