table of Contents
2 Simple fully connected neural network
3 Simple to add a network layer:
References: https://www.cnblogs.com/geeksongs/p/13204568.html
1 Introduction:
When we use tensorflow, if we can't use functional api for programming, then some complex neural network structures will not be realized, and we can only use simple one-way models to stack layer by layer . If it is a little more complicated and encounters a neural network with residual modules like Resnet, it is impossible to stack this kind of network with a simple neural network stacking method. Let's use functional API to write a simple fully connected neural network :
2 Simple fully connected neural network
As shown
Benefits are:
Multi-input model
For example, to judge whether two are the same kind
Multiple input multiple output model
3 Simple to add a network layer:
This part will be used all the time, then reference materials will be enough
References:
https://www.cnblogs.com/geeksongs/p/13204568.html
TF2.0 actual combat tutorial of station b Sunyueguanghua