新手玩家
- 贡献度
- 0
- 金元
- 235
- 积分
- 24
- 精华
- 0
- 注册时间
- 2019-3-21
|
One of the most common forms of an ANN is the multilayer perceptron, as shown in Fig. 1. The big circles represent the processing units, called neurons, and the interconnects between them are called synapses. The synapses each carry a weighted value that updates (learns) as the network is trained. The number of synapses and neurons depends entirely on the application. For photonic modelling, the input layer has N1 neurons: the number of user-specified variable device parameters. And the output layer has NL neurons: the number of user-specified device outputs, where L is the number of layers in the network. The number of hidden layers and the number of neurons in those layers remain an optimization problem on their own, though one layer is enough to generalize any nonlinear input–output relationship to a reasonable degree of accuracy
|
|
[复制链接]
发表于 2019-3-18 22:24
