A.Generally speaking, within 5 layers of Sigmoid network, the gradient will degenerate to 0, making it difficult to train. This phenomenon is called the vanishing gradient phenomenon.

B.Sigmoid function is monotonous and continuous, easy to derive, bounded output, and the network is easy to converge.

C.Sigmoid output is not centered on 0

D.When the network is very deep, the modulus of the gradient becomes smaller and smaller

A.Iterate until convergence

B.Solve equations

C.Initialization parameters

D.E-step and M-step

A.mox.file .copy

B.mox.file .rename

C.mox.file .remove

D.mox.file .copyparallel

A.Naive Bayes

B.Linear regression

C.TextCNN

D.SVM

A.Use a circle to frame the target

B.Enter a category label for the image

C.Use an ellipse to frame the target

D.Use a rectangle to frame the target