
上QQ阅读APP看书,第一时间看更新
There's more...
There are three main neural network architectures in neural networks:
- Feedforward ANN: This is a class of neural network models where the flow of information is unidirectional from input to output; thus, the architecture does not form any cycle. An example of a Feedforward network is shown in the following image:

Feedforward architecture of neural networks
Feedback ANN: This is also known as the Elman recurrent network and is a class of neural networks where the error at the output node used as feedback to update iteratively to minimize errors. An example of a one layer Feedback neural network architecture is shown in the following image:

xFeedback architecture of neural networks
Lateral ANN: This is a class of neural networks between Feedback and Feedforward neural networks with neurons interacting within layers. An example lateral neural network architecture is shown in the following image:

Lateral neural network architecture