ehsanhaghighat / sciann

Deep learning for Engineers - Physics Informed Deep Learning
htttp://sciann.com
Other
325 stars 116 forks source link

Res-Net Structure #101

Closed YunhaoYuan closed 8 months ago

YunhaoYuan commented 11 months ago

Dear SciANN community,

I have noticed that the funtional supports the res-net structure by simply set res_net argument to True, but did not explain how to modify the shortcut connections. (res_net: (True, False). Constructs a resnet architecture. Defaulted to False.)

I am wondering how to define a network with multiple residual blocks with different hidden layers and designs.

For example, I want to build a network with the following design: number of input: 1 residual block 1: 2 layers, with [30, 30] hidden neurons residual block 2: 3 layers, with [40, 30, 20] hidden neurons number of output: 2

Any suggestions / examples would be highly appreciated.

Thank you!