You are on page 1of 3

Optimization

Algorithms

Gradient descent
deeplearning.ai
with momentum
Gradient descent example

deeplearning.ai
Andrew Ng
Implementation details
Oniteration 𝑡 :
 

Compute on the current mini-batch


 

 
𝑣 𝑑𝑊 = 𝛽 𝑣 𝑑𝑊 + ( 1− 𝛽 ) 𝑑𝑊
 
𝑣 𝑑𝑏 =𝛽 𝑣 𝑑𝑏 + ( 1− 𝛽 ) 𝑑𝑏
 
𝑊 =𝑊 − 𝛼 𝑣 𝑑𝑊 ,  𝑏=𝑏 − 𝛼 𝑣 𝑑𝑏

 
Hyperparameters:  
𝛽=0.9
Andrew Ng

You might also like