Introduction to Neural Networks with Python

dataHacker.rs
Introduction to
Neural Networks

dataHacker.rs
NN Intro
 Logistic Regression
 Forward Propagation
 Cost Function
 Backward Propagation
 Neural Network
 Brain Analogy
 Logistic Regression Implementation

dataHacker.rs
Logistic Regression
 Supervised learning algorithm
 Binary classification problem
 Labels: 0 or 1
 Example: cat picture

dataHacker.rs
Logistic Regression
Linear function
 Parameters: w, b
Sigmoid Activation function
 Range 0-1
“weights”
“bias vector”
“input data”
“output”

dataHacker.rs
Simple Neuron
Logistic regression
 One neuron – two functions
 1 -Linear function
 2 - Activation function

dataHacker.rs
Computation Graph
Forward propagation & Backward propagation
 Cost Function
 Measures how well our
algorithm is working

dataHacker.rs
Key components
Three key components to train Neural Networks
 1. Score function (Linear+Activation)
 2. Loss Function
 3. Optimization

dataHacker.rs
Forward propagation
“input data”
“weight”
“bias” Linear function Activation function Loss function
Logistic Regression

dataHacker.rs
Cost Function
How our algorithm
works?
Visualizing the cost
function

dataHacker.rs
Backward propagation
 Optimization
 Gradient descent
 Learning rate
What is the correct value
of parameters?

dataHacker.rs
Backward propagation
 Gradient of L with respect to parameters
 Calculate derivatives
 Local gradient X Upstream gradient

dataHacker.rs
Neural Network
Input layer
Hidden layers
Output layer

dataHacker.rs
Neural Network
Logistic Regression
Neural Network with 5
hidden layers
“shallow”
“shallow” “shallow”
“deep”
Neural Network with single
hidden layer
Neural Network with 2
hidden layers

dataHacker.rs
Logistic Regression
Implementation

Introduction to Neural Networks with Python

In this document

More Related Content

What's hot

Similar to Introduction to Neural Networks with Python

More from dataHacker. rs

Recently uploaded

Introduction to Neural Networks with Python