Advertisement
Guest User

Untitled

a guest
Jun 20th, 2019
85
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 0.23 KB | None | 0 0
  1. # dout: ∂L/∂o
  2. # a: the input of FC layer
  3. # the shape of a: (batch size, hidden layer neuron)
  4. # the shape of dw: (hidden layer neuron, num of classes)
  5. dw = a.reshape(a.shape[0], -1).T.dot(dout)
  6. db = np.sum(dout, axis=0)
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement