cs621: artificial intelligence lecture 13: region counting; perceptron training

14
CS621: Artificial Intelligence Lecture 13: region counting; perceptron training Pushpak Bhattacharyya Computer Science and Engineering Department IIT Bombay

Upload: finn-ayala

Post on 30-Dec-2015

29 views

Category:

Documents


0 download

DESCRIPTION

CS621: Artificial Intelligence Lecture 13: region counting; perceptron training. Pushpak Bhattacharyya Computer Science and Engineering Department IIT Bombay. Number of regions founded by n hyperplanes in d-dim passing through origin is given by the following recurrence relation - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

CS621: Artificial IntelligenceLecture 13: region counting;

perceptron training Pushpak Bhattacharyya

Computer Science and Engineering DepartmentIIT Bombay

Page 2: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Number of regions founded by n hyperplanes in d-dim passing through origin is given by the following recurrence relation

we use generating function as an operating function

Boundary condition:

1 hyperplane in d-dim

n hyperplanes in 1-dim,

Reduce to n points thru origin

The generating function is

1,1,, 1 dnddn RRR n

2

2

1,

,1

n

d

R

R

d

n d

ndn yxRyxf

1 1

,),(

Page 3: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

22 2

,1

2

1,1

2 2

,1

2 2

1,11

1 1

1,

2 1

,1

1 1

1,

1 1

,

2

),(

),(

),(

),(

n

n

n d

dndn

n

nn

n d

dndn

d

n d

ndn

d

n d

ndn

d

n d

ndn

n d

dndn

d

n d

ndn

yxyxR

yxRyxRyxfx

yxRyxRyxfxy

yxRyxRyxfx

yxRyxf

Page 4: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

After all this expansion,

since other two terms become zero

xyyxyxyxyxR

xyRyxRxyRyxR

yxRyxf

n

n

d

dd

n d

ndn

n

nn

d

dd

d

n d

ndn

d

n d

ndn

222

),(

112 2

,

1,1

1

1,

1

,1

2 2

,

1 1

,

1

121

2 2

1,1,1,

2

2222

)(

d

d

d

d

nn

d

d

n d

ndndndn

yx

yxxyxyxy

yxRRR

),(),(),( yxfxyyxfxyxf

Page 5: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

This implies

also we have,

Comparing coefficients of each term in RHS we get,

].....)1(...)1()1(1[

]........[2

2)]1(1[

1),(

2),(]1[

22

32

1

1

dd

d

d

d

d

d

yxyxyx

yyyyx

yxyx

yxf

yxyxfxyx

d

n d

ndn yxRyxf

1 1

,),(

Page 6: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

1

0

12d

i

n

iC

Comparing co-efficients we get

dnR ,

Page 7: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Applied To Perceptron

• No. of dimensions = (m+1)• No. of hyperplanes = 2m

)1(,2,

mdn mRR

m

iiC

m

0

)12(2

Page 8: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Calculating Upper Bound On Total Threshold Functions (1/4)

m

iimCR

m

m

0

)12(

)1(,22

]2...2221[2

]!

2...

!3

2

!2

2

!12

1[2

]...1[2

2

2

32

32

)12(2

)12(1

)12(

mmmm

mmmm

m

m

CCCmmm

Page 9: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Upper bound contd. (2/4)

]2...2221[2232

)1(,2

mmmm

mmR

Using summation of GP formula,

]12

12[2

)1(

)1(,2

m

m

m

m

mR

Page 10: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Upper bound contd. (3/4)

12

122

)1(

)1(,2 m

m

m

m

mR

)2(

2

2

22

12

122

2

2

2

2

1

m

m

m

mm

m

mm

O

Page 11: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Upper bound contd. (4/4)

• No. of hyperplanes = n = 2m

• No. of threshold functions = R(n,d) = )2(2mO

Page 12: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Functional Mapping of Brain

Page 13: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

The human brain

Pre-frontal cortex: Self control

Insula:Intuition &empathy

Anterior CingulateCortex:Anxiety

Hippocampus:Emotional memory

Hypothalamus:Control of HormonesPituitary gland:

Mother instincts

Amygdala:Aggresion

Page 14: CS621: Artificial Intelligence Lecture 13: region counting; perceptron training

Functional Map of the Brain