Keras在用于反向传播的多个输出的多个损耗中找到最佳损耗

Question

I am using a multiple output model in keras for a regression problem.

For example the model is like :

model = Model(inp, [out1, out2, out3])

I want to calculate loss of each output but based on each loss, select one and backpropagate only through that and make every other weights 0. It is like a loss that finds the best loss and output among all loss and learns only through that output.

For example, out 1 loss was minimum, so my weight will be [1.0, 0.0, 0.0],so that model will not learn anything from out2 and out3.

size of out1, out2 and out3 are same.

I am not so experienced with keras, is there any way to do it.

Thanks for the help.

Answer 1

多个输出模型记录在： https : //keras.io/getting-started/functional-api-guide/#multi-input-and-multi-output-models