Introduction to Machine Learning – Perceptron Learning Algorithm – Neural Networks

View Github Repository Open presentation in a new window

abisz

See all presentation from abisz

Introduction to Machine Learning – Perceptron Learning Algorithm – Neural Networks

0 1

talk-ml-introduction

An introduction talk about Machine Learning

On Github abisz / talk-ml-introduction

Introduction to Machine Learning

Simon Reinsperger | piedcode.com | @simon_rsp https://github.com/abisz/talk-ml-introduction

What is Machine Learning?

"Science of getting computers to act without being explicitly programmed" - Andrew Ng, Standford

Regular Programming:

Input + Rules => Output

Machine Learning:

Input + Output => Rules

Difference to AI and Deep Learning

https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/ Artificial Intelligence Started in the 50s Rule-based No learning Example TicTacToe - unbeatable Hype during the last years, new hardware, faster processing Machine Learning Started in the 80s Subcategory of AI - ML is always AI, AI is not always ML Learn from existing data to make predictions about unknown data No handcoded rules necessary Deep Learning New trend in ML Inspired by the brain Many small neurons ("stupid/simple") in a network Surprise Extremely good performance in image/language recognition Unknown what's going on.

When to use Machine Learning?

  • a pattern exists
  • cannot be solved mathematically
  • data is available

Types of Machine Learning

  • Supervised Learning
    • input/output pairs
    • classification
  • Unsupervised Learning
    • only input
    • finding clusters/categories
  • Reinforcement Learning
    • input + grade of output
    • finding ideal behaviour for context

Examples

Perceptron Learning Algorithm

  • Supervised ML
  • Classification
  • Data needs to be linearly separable

Let's see it in action!

abisz.github.io/perceptron-example/

Neural Networks

  • Multiple perceptrons
  • Output from one (hidden) layer -> input for next

Tensorflow Playground

playground.tensorflow.org

Convolutional Network

http://joelouismarino.github.io/blog_posts/blog_googlenet_keras.html

Image Recognition

http://www.auduno.com/2016/06/18/peeking-inside-convnets/

Not a trivial task...

http://imgur.com/gallery/TTpIGvo
http://www.boredpanda.com/dog-food-comparison-bagel-muffin-lookalike-teenybiscuit-karen-zack/

Thank you for listening

http://giphy.com/gifs/thanks-thank-you-kazoo-kid-l3q2wJsC23ikJg9xe
Introduction to Machine Learning Simon Reinsperger | piedcode.com | @simon_rsp https://github.com/abisz/talk-ml-introduction