Cost Function and Backpropagation

Cost Function : article

Neural Network (Classification)

L, sl, K:

• L = total number of layers in the network
• sl = number of units (not counting bias unit) in layer l
• K = number of output units/classes

EX:

• L: 總共的 layer 數

   L = 4
  

• Sl: 每一 layer 上的 unit(Neuron) 數

   s1 = 3, s2 = 5, s4 = sL = 4
  

Binbary classification & Multi-class classification (K classes)

Binary classification	Multi-class classification (K classes)
y = 0 or 1	y = ℝ^K
1 output unit	K output units
sL = 1 (K = 1)	sL = k (k ≥ 3)

• 當 K 超過2種以上就要用 Multi-calss classification

Cost function

• Logistic regression:

• Neural network: