intro to neural networks

55
Intro to Neural Networks Supervised Learning: Perceptrons and Backpropagation

Upload: galia

Post on 23-Feb-2016

62 views

Category:

Documents


0 download

DESCRIPTION

Supervised Learning: Perceptrons and Backpropagation. Intro to Neural Networks. Neural Network ==. Connectionist /ism== Parallel Distributed Processing (PDP). Neural Networks assume . Intelligence is emergent. 1943 - McCullough Pitts Artificial Neuron. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Intro to Neural Networks

Intro to Neural Networks

Supervised Learning: Perceptrons and Backpropagation

Page 2: Intro to Neural Networks

Neural Network ==

Connectionist /ism==Parallel Distributed Processing (PDP)

Page 3: Intro to Neural Networks

Neural Networks assume

Intelligence is emergent

Page 4: Intro to Neural Networks

1943 - McCullough Pitts Artificial Neuron

Page 5: Intro to Neural Networks

1943 - McCullough Pitts Artificial Neuron

Page 6: Intro to Neural Networks

Perceptron Learning 1958

Page 7: Intro to Neural Networks

Perceptron Learning 1958

Page 8: Intro to Neural Networks

Perceptron Learning 1958

Page 9: Intro to Neural Networks

Perceptron Learning 1958

Page 10: Intro to Neural Networks

Perceptron Learning 1958

Page 11: Intro to Neural Networks

Linear Seperability Problem 1965

Page 12: Intro to Neural Networks

Linear Seperability Problem 1965

Page 13: Intro to Neural Networks

Backpropagation

Used to train multilayer feedforward networks

Page 14: Intro to Neural Networks

Backpropagation

Page 15: Intro to Neural Networks

Backpropagation

Used to train multilayer feedforward networks

Assumes a continuous activation function

Page 16: Intro to Neural Networks

Backpropagation - Activation

Page 17: Intro to Neural Networks

Backpropagation

Used to train multilayer feedforward networks

Assumes a continuous activation function

Delta rule

Page 18: Intro to Neural Networks

Backpropagation Delta rulePerceptron update rule was:

Backprop update rule is:

Δw = c(desired − sign(actual))x

Δw = c(error)x

Page 19: Intro to Neural Networks

Backpropagation Delta ruleError of an output node:

error j = (1−output j2)(desired j − actual j )

Page 20: Intro to Neural Networks

Backpropagation Delta ruleError of a hidden node:

errori = (1−outputi2)( error j *wij

j∑ )

Page 21: Intro to Neural Networks

Backpropagation Delta rule

Page 22: Intro to Neural Networks

Backpropagation Delta rule

Page 23: Intro to Neural Networks

Backpropagation Delta rule

Page 24: Intro to Neural Networks

Backpropagation Delta rule

Page 25: Intro to Neural Networks

Backpropagation

demo

Page 26: Intro to Neural Networks

Inductive Bias

Page 27: Intro to Neural Networks

Inductive Bias

Encoding / Feature Extraction# neurons used# layers usedOutput mapping

Page 28: Intro to Neural Networks

Domains

Classification

Page 29: Intro to Neural Networks

Domains

ClassificationPattern Recognition

Page 30: Intro to Neural Networks

Domains

ClassificationPattern RecognitionContent Addressable Memory

Page 31: Intro to Neural Networks

Domains

ClassificationPattern RecognitionContent Addressable MemoryPrediction

Page 32: Intro to Neural Networks

Domains

ClassificationPattern RecognitionContent Addressable MemoryPredictionOptimization

Page 33: Intro to Neural Networks

Domains

ClassificationPattern RecognitionContent Addressable MemoryPredictionOptimizationFiltering

Page 34: Intro to Neural Networks

The good

Degrade gracefully

Page 35: Intro to Neural Networks

The good

Degrade gracefullySolve ill-defined problems

Page 36: Intro to Neural Networks

The good

Degrade gracefullySolve ill-defined problemsFlexible

Page 37: Intro to Neural Networks

The good

Degrade gracefullySolve ill-defined problemsFlexibleGeneralization

Page 38: Intro to Neural Networks

The bad

Time & Memory

Page 39: Intro to Neural Networks

The bad

Time & MemoryBlack box

Page 40: Intro to Neural Networks

The bad

Time & MemoryBlack boxTrial and Error

Page 41: Intro to Neural Networks

When not to use Feedforward net If you can draw a flow chart or

formula

Page 42: Intro to Neural Networks

When not to use Feedforward net If you can draw a flow chart or

formula If a piece of hardware or software

already exists that does what you want

Page 43: Intro to Neural Networks

When not to use Feedforward net If you can draw a flow chart or

formula If a piece of hardware or software

already exists that does what you want

If you want to functionality to evolve

Page 44: Intro to Neural Networks

When not to use Feedforward net If you can draw a flow chart or

formula If a piece of hardware or software

already exists that does what you want

If you want to functionality to evolvePrecise answers are required

Page 45: Intro to Neural Networks

When not to use Feedforward net If you can draw a flow chart or

formula If a piece of hardware or software

already exists that does what you want

If you want to functionality to evolvePrecise answers are requiredThe problem could be described in a

lookup table

Page 46: Intro to Neural Networks

When to use feedforward netYou can define a correct answer

Page 47: Intro to Neural Networks

When to use feedforward netYou can define a correct answerYou have a lot of training data with

examples of right and wrong answers

Page 48: Intro to Neural Networks

When to use feedforward netYou can define a correct answerYou have a lot of training data with

examples of right and wrong answers

You have lots of data but can’t figure how to map it to output

Page 49: Intro to Neural Networks

When to use feedforward netYou can define a correct answerYou have a lot of training data with

examples of right and wrong answers

You have lots of data but can’t figure how to map it to output

The problem is complex but solvable

Page 50: Intro to Neural Networks

When to use feedforward netYou can define a correct answerYou have a lot of training data with

examples of right and wrong answers

You have lots of data but can’t figure how to map it to output

The problem is complex but solvableThe solution is fuzzy or might change

slightly

Page 51: Intro to Neural Networks

Examples2007 Rechnender Raum’s

Inverted Machine

Page 52: Intro to Neural Networks

Examples

Jonathan McCabe’sNervous States 2006Each pixel is the Output state of aNeural network givenDifferent inputs

Page 53: Intro to Neural Networks

Examples

2007 Phillip StearnsAANN: Artificial Analog Neural Network

Page 54: Intro to Neural Networks

Examples

Page 55: Intro to Neural Networks

Examples

Ted?