Lesson 17 Network Architecture for Deep Learning
Network for Deep Learning should have a lot of artificial neurons, which are connected like a net, including convolution or pooling layers. As a result, it should have multiple layers. The following is an example.
Here, the first step is a convolution process labeled C, the second one is a pooling process labeled P, and then a complex network with artificial neurons. Finally, we get two outputs.
Layer for convolution process is called Convolution Layer, and layer for pooling process is called Pooling Layer. Each layer which consists of multiple artificial neurons is called Full Connected Layer. When we work on Deep Learning, we will create many suitable layers for the question you study. The probability of getting correct answer depends on the number of layers. But you may need more time for the repeated computation, and also get no appropriate parameter after the calculation.