Building blocks of
Deep Learning
Keshan Sodimana
Today’s menu
● What is a deep neural network
● How forward propagation works
● What are activation, cost function and regularization
● Gradient descent and optimizations
● Backpropagation
● Let’s practice….
BUT WAIT...
Do I need to have a good
knowledge in Mathematics? I
have done nothing on Maths
since my ALs.
y = mx + c
z = f(x)
∑f(x)
dx/dy
Do you remember?
What is a deep neural network
Forward propagation
Activation functions
Regularization
Cost function
Our goal is to minimize the cost.
Gradient descent
Finding the global minimum
Optimization
Backpropagation
Spoiler Alert : This could be the hardest part to digest
Backpropagation
Remember : back
propagation is used to
calculate the gradient
of the loss function with
respect to the
parameters
Can you remember the chain rule in calculus?
Ok let’s go for a test run...

Building blocks of deep learning

Editor's Notes

  • #16 Note to myself: iPython notebook needs to be prepared