TensorFlow APIs
When using TensorFlow, you can choose from the sequential, functional, or subclassing APIs to define your models. For most, the sequential API will be the go-to option. However, as time goes by and you are exposed to more complexity, your needs will expand as well.
The sequential API is the simplest API used for creating TensorFlow models. It works by stacking different layers one after the other. For example, you will create a sequential model with a first layer that's a convolution layer, followed by a dropout layer, and then a fully connected layer. This model is sequential as the input data will be passed to each defined layer sequentially.
The functional API provides more flexibility. You can define models with different layers that interact with each other not in a sequential manner. For instance, you can create two different layers both of which will feed into a third one. This can be easily achieved with the functional API.
Model subclassing allows...