Introduction
to
Machine Learning
Tyson Thomas
tysonthomas.90@gmail.com
About Me
Currently Software Architect at
Kadima Consulting
BE ECE - CMRIT
2008-12
Co-Founder of
"Machine Learning is the field of study that gives computers the ability to learn without being explicitly programmed."
- Arthur Samuel
Types of Machine Learning
Applications
- Detecting spam emails
- Forecasting stock prices
- Recognizing objects in an image
- Diagnosis of illness
- Self-driving cars
Deep Learning
Neural Networks
Vaugley mimics the process of how the brain operates with neurons that fire bits of information
Neural network that recognizes images of a cat
Example Classification Problem
Example Classification Problem - Solution
Classification Problem - Mathematical Representation
A Neuron classifying a data point into two kinds
Mathematical Representation of a single neuron
A neuron classifies any data point into one of two kinds
Using a single neuron to perform image recognition
28 Pixels * 28 Pixels = 784 Total Pixels
Training Dataset
Classification of an image as 8 or not?
Weights Distribution
How are the parameters determined?
(weights & biases)
- Backpropagation
- Gradient Descent
Deep Neural Networks
Classification Problem - 2
Classification Problem - 2
Feature Space Representation
A hidden layer transforms inputs to feature space, making it linearly classifiable
Neural network can extract insights from (seemingly) random signals
Classification Problem - 3
Classification Problem - 3
Deep Convolutional Neural Networks
Real World
Examples
Self Driving Car in Arizona, USA
Shopping Technology
Google DeepMind
Alpha Go
Alpha Go Zero using
Reinforcement Learning
Google Deep Mind reduces energy cost in datacenter
Locomotion Behaviours in Rich Environments
Getting into Machine Learning
Raspberry Pi + Intel Movidius Neural Chip
Nvidia Jetson
Resources
Questions?
Thank you
Tyson Thomas
tysonthomas.90@gmail.com
Machine Learning
By Tyson Thomas
Machine Learning
- 709