delta 2

Information about delta 2

Published on January 9, 2008

Author: Simeone

Source: authorstream.com

Content

The delta rule:  The delta rule Learn from your mistakes:  Learn from your mistakes If it ain’t broke, don’t fix it.:  If it ain’t broke, don’t fix it. Outline:  Outline Supervised learning problem Delta rule Delta rule as gradient descent Hebb rule Supervised learning:  Supervised learning Given examples Find perceptron such that Example: handwritten digits:  Example: handwritten digits Find a perceptron that detects “two”s. Delta rule:  Delta rule Learning from mistakes. “delta”: difference between desired and actual output. Also called “perceptron learning rule” Two types of mistakes:  Two types of mistakes False positive Make w less like x. False negative Make w more like x. The update is always proportional to x. Objective function:  Objective function Gradient update Stochastic gradient descent on E=0 means no mistakes. Perceptron convergence theorem:  Perceptron convergence theorem Cycle through a set of examples. Suppose a solution with zero error exists. The perceptron learning rule finds a solution in finite time. If examples are nonseparable:  If examples are nonseparable The delta rule does not converge. Objective function is not equal to the number of mistakes. No reason to believe that the delta rule minimizes the number of mistakes. Memorization & generalization:  Memorization & generalization Prescription: minimize error on the training set of examples What is the error on a test set of examples? Vapnik-Chervonenkis theory assumption: examples are drawn from a probability distribution conditions for generalization contrast with Hebb rule:  contrast with Hebb rule Assume that the teacher can drive the perceptron to produce the desired output. What are the objective functions? Is the delta rule biological?:  Is the delta rule biological? Actual output: anti-Hebbian Desired output: Hebbian Contrastive Objective function:  Objective function Hebb rule distance from inputs Delta rule error in reproducing the output Supervised vs. unsupervised:  Supervised vs. unsupervised Classification vs. generation I shall not today attempt further to define the kinds of material [pornography] … but I know it when I see it. Justice Potter Stewart Smooth activation function:  Smooth activation function same except for slope of f update is small when the argument of f has large magnitude. Objective function:  Objective function Gradient update Stochastic gradient descent on E=0 means zero error. Smooth activation functions are important for generalizing the delta rule to multilayer perceptrons.:  Smooth activation functions are important for generalizing the delta rule to multilayer perceptrons.

Related presentations


Other presentations created by Simeone

Introduction to Design Pattern
14. 02. 2008
0 views

Introduction to Design Pattern

Erikson and Horney
15. 01. 2008
0 views

Erikson and Horney

Sudden Illnesses 9
14. 01. 2008
0 views

Sudden Illnesses 9

Customer loyalty presentaion
15. 03. 2008
0 views

Customer loyalty presentaion

topic5 ravi
10. 01. 2008
0 views

topic5 ravi

cloning
15. 01. 2008
0 views

cloning

aecma
16. 01. 2008
0 views

aecma

Tox Made Simple
18. 01. 2008
0 views

Tox Made Simple

NEO YOUNG CREATIVE AWARD 2007
04. 02. 2008
0 views

NEO YOUNG CREATIVE AWARD 2007

GLJ Mobile Logistics121207
04. 02. 2008
0 views

GLJ Mobile Logistics121207

CohenLevesque
07. 02. 2008
0 views

CohenLevesque

pring
11. 02. 2008
0 views

pring

Ancient Cosmetic toxicology
10. 01. 2008
0 views

Ancient Cosmetic toxicology

1stReviewppt
25. 01. 2008
0 views

1stReviewppt

newsfile469 1
28. 01. 2008
0 views

newsfile469 1

15 Comets
16. 01. 2008
0 views

15 Comets

imci c slides
25. 02. 2008
0 views

imci c slides

Friday
28. 02. 2008
0 views

Friday

Safetyprogram
05. 03. 2008
0 views

Safetyprogram

2 20071224184840
11. 03. 2008
0 views

2 20071224184840

websearch
19. 03. 2008
0 views

websearch

schreffler hopkins
21. 03. 2008
0 views

schreffler hopkins

CKtrain green
15. 01. 2008
0 views

CKtrain green

era2
03. 04. 2008
0 views

era2

mother100703
09. 01. 2008
0 views

mother100703

dahlbom
21. 01. 2008
0 views

dahlbom

Comics and Mangas
18. 02. 2008
0 views

Comics and Mangas

AMC Online Trip Listings
06. 02. 2008
0 views

AMC Online Trip Listings

CPZvezda Brus
22. 01. 2008
0 views

CPZvezda Brus