8/14/2019 Recogniton Factor

1/38

The Recognition Factor

8/14/2019 Recogniton Factor

2/38

1. Practical Applications

2. Distributed Knowledge

3. Patterns and Clustering4. Pattern Recognition

5. Reliable Networks

8/14/2019 Recogniton Factor

3/38

You already know a

lot of this stuffWhat I am trying to

do is to get you to

see it differently,more clearly

Then you will see

things in everydayknowledge differently

than you did before

1

8/14/2019 Recogniton Factor

4/38

Spot thePlanets

8/14/2019 Recogniton Factor

5/38

Spot the Planets

8/14/2019 Recogniton Factor

6/38

The Theory

To teach is

to model and to demonstrate

To learn is

to practice and to reflect

Pretty simple, eh?

No cheats, no shortcuts

8/14/2019 Recogniton Factor

7/38

To model what?

To practice what?

That is the topic of this talk

8/14/2019 Recogniton Factor

8/38

For example

Evaluate [6 - (5 - 7(7 - 3) + 5)] + 4

2. -33

3. 28

4. 215. 13

To teach the concept of brackets, would you use this

same example over and over? Of course not.

Why not?

8/14/2019 Recogniton Factor

9/38

Because you are trying to teach aconcept, not a fact

And the concept is something

deeperthan what you see in anygiven example

Fair enough

http://classes.aces.uiuc.edu/ACES100/Mind/c-m2.html

http://classes.aces.uiuc.edu/ACES100/Mind/c-m2.htmlhttp://classes.aces.uiuc.edu/ACES100/Mind/c-m2.html

8/14/2019 Recogniton Factor

10/38

But is the concept best

thought of as:

A rule?

A pattern?

8/14/2019 Recogniton Factor

11/38

2Representation

treestands for

8/14/2019 Recogniton Factor

12/38

stands for?

Or is caused by?

Distributed Representation

= a pattern of connectivity

8/14/2019 Recogniton Factor

13/38

8/14/2019 Recogniton Factor

14/38

The theoryConcepts are not words

They arepatterns in a network

(like the mind, like society)There is no specificplace the concept is located it

is distributed as a set of connections across the

network

Other concepts are embedded in the same network

they form parts of each other,

they effect each other

8/14/2019 Recogniton Factor

15/38

3Self-organizing systems acquire new structure without specific

interference from the outside. They exhibit qualitative macroscopic

changes such as bifurcations or phase transitions.

http://www.christianhubert.com/hypertext/self_organization.html

http://www.christianhubert.com/hypertext/self_organization.htmlhttp://www.christianhubert.com/hypertext/self_organization.html

8/14/2019 Recogniton Factor

16/38

The way things connect is reflective

of the properties of those things

8/14/2019 Recogniton Factor

17/38

They obey

the laws of

physics

(Force patterns inconstruction

http://paginas.ufm.

)

http://paginas.ufm.edu/arquitemas/ffconclusions03.htmlhttp://paginas.ufm.edu/arquitemas/ffconclusions03.html

8/14/2019 Recogniton Factor

18/38

They are influenced by

external stimuli

http://www.williamcalvin.com/1990s/1995Handbook.htm

http://www.williamcalvin.com/1990s/1995Handbook.htmhttp://www.williamcalvin.com/1990s/1995Handbook.htm

8/14/2019 Recogniton Factor

19/38

Scale-free

networks andpower laws

are

just one type

of network

where earlylinks are

attractors

8/14/2019 Recogniton Factor

20/38

Different kinds of networks detect different kinds of patterns

http://neural.cs.nthu.edu.tw/jang/courses/cs5652/lippman.gif

http://neural.cs.nthu.edu.tw/jang/courses/cs5652/lippman.gifhttp://neural.cs.nthu.edu.tw/jang/courses/cs5652/lippman.gif

8/14/2019 Recogniton Factor

21/38

4

We are natural pattern recognizers

thats what our brains do

hierarchical neural network for visual pattern recognition

8/14/2019 Recogniton Factor

22/38

Some things (like edge detection) we

do because of the way were wired

8/14/2019 Recogniton Factor

23/38

For mostthings, though,

there is more

at work

http://www.mcs.drexel.edu/~gcmastra/strange2.html

http://www.mcs.drexel.edu/~gcmastra/strange2.htmlhttp://www.mcs.drexel.edu/~gcmastra/strange2.html

8/14/2019 Recogniton Factor

24/38

What is it?

8/14/2019 Recogniton Factor

25/38

Duck Rabbit

8/14/2019 Recogniton Factor

26/38

Attractors = the tendency of the network tointerpret a phenomenon one way as

opposed to another

(energy states of variousneural net configurations)

Associative memory =pattrerns of connectivity =

the creation of attractors =

recognition

http://7ka.mipt.ru/~yevin/vismath/

http://7ka.mipt.ru/~yevin/vismath/http://7ka.mipt.ru/~yevin/vismath/

8/14/2019 Recogniton Factor

27/38

Knowledge is like recognitionLearningis likeperception

the acquisition of new patterns of

connectivity

through experience

8/14/2019 Recogniton Factor

28/38

Like I said, you already know this phenomenon,

youve already seen it

Emergent Learninghttp://growchangelearn.blogspot.com/2007/02/emergent-learning.html

Tom Haskins

"Now I get it"

A-ha!

"Out of the blue"

"My mind leaped""Did an about-face"

"Shut up and did it"

Sudden breakthrough

http://growchangelearn.blogspot.com/2007/02/emergent-learning.htmlhttp://growchangelearn.blogspot.com/2007/02/emergent-learning.html

8/14/2019 Recogniton Factor

29/38

Knowledge is

recognition

Its a belief you

cant not have

Like after youvefound Waldo

8/14/2019 Recogniton Factor

30/38

5Pattern Recognition

8/14/2019 Recogniton Factor

31/38

http://www.sund.de/netze/applets/BPN/bpn2/ochre.html

Pattern recognition is based on similarity

between the current phenomenon and

previously recognized phenomena

http://www.sund.de/netze/applets/BPN/bpn2/ochre.htmlhttp://www.sund.de/netze/applets/BPN/bpn2/ochre.html

8/14/2019 Recogniton Factor

32/38

What we wantis for students to

recognize patterns in existing networks in communities of experts,

communities of practice

Thats why we model and demonstrate

8/14/2019 Recogniton Factor

33/38

But what kindof

network do we

want to model for

our students?

For that matter,what kind of

network do we

want for

ourselves?

To maximize

knowledge?

To little connection and information never propagates

Too much connection and

information propagates too

quickly

8/14/2019 Recogniton Factor

34/38

The internet itself illustrates a sound set of

principles, grounded by two majorcharacteristics: simple services with realistic

scope. "Simple service or simple devices with

realistic scope are usually able to offer a

superior user experience compared to acomplex, multi-purpose service or device".

8/14/2019 Recogniton Factor

35/38

Effective networks are

Decentralized

Distributed

Disintermediated

Disaggregated

Dis-Integrated

Democratic

Dynamic

Desegregated

8/14/2019 Recogniton Factor

36/38

Democratic =

The Semantic Condition

Reliable networks support

Autonomy

Diversity

Openness Connectivity

8/14/2019 Recogniton Factor

37/38

How is this practical?

Ask yourself

To teach the concept of brackets, would you use thissame example over and over? Of course not.

Why not?

Because of the need fordiversity.

Diverse experiences create better networks

than monotonous experiences

8/14/2019 Recogniton Factor

38/38

Thank you

http://www.downes.ca