李宏毅hw1_covid19预测_代码研读+想办法降低validation的loss(Kaggle目前用不了)
1.考虑调整这个neuralnetwork的结构尝试让这个loss降低(1)Linear(inputdim,64)-ReLU-Linear(64,1),loss=0.7174(2)Linear(inputdim,64)-ReLU-Linear(64,64)-ReLU-Linear(64,1),loss=0.6996(3)这样设置的话,3个linear层中间夹着2个ReLU激活函数:(4)2个Lin