Machine Learning

geckokittenΤεχνίτη Νοημοσύνη και Ρομποτική

17 Οκτ 2013 (πριν από 4 χρόνια και 22 μέρες)

123 εμφανίσεις

Machine Learning

Hilary Hamlin

Jacob Curtis

H
istory


Arthur Samuel (1959)

o
"Field of study that gives computers the ability to
learn without being explicitly programmed"



Tom
M.Mitchell

(1998)

o
"A computer program is said to learn from
experience E with respect to some class of tasks T
and performance measure P, if its performance at
tasks in T, as measured by P, improves with
experience E"



Introduction


Learning Machines = Programs that learn



Create Knowledge Bases

o
Symbolic

o
Analog



Learn Concepts

o
Supervised

o
Unsupervised


Supervised Learning


Labeled Training Data

o
pairs of input & output


Generalizes


Determine Labels for new Info

INPUT


SYSTEM

OUTPUT

Training Examples

ƒ

{ (x1, f(x1)), (x2, f(x2))...(
xn,f
(
xn
)) }


for some unknown function (system) y = f(x)

ƒFind f(x)



ƒPredict y'=f(x')




OUTPUT

INFERRED FUNCTION

NEW DATA

INPUT

ID3

BATGIRL

Checkers


world's first self
-
learning program




limited amount of available computer
memory


o
alpha
-
beta pruning


o
rote learning


alpha
-
beta pruning

A

B

C

D

E

F

G

H

I

J

K

MAX

MIN

( 3 )

(12)

( 5)

( 3 )

( 9)

( 2 )

( 7 )

<=3



>=3

<=5


3

<=3

<=2

Min
-
Max O(
b^m
)



Alpha
-
Beta O(b^(m/2))


Checkers: Pruning Tree


Unsupervised Learning


N
o
training data set


N
o
clear 'right' output


Finds
patterns in
data and labeling accordingly


Commonly implemented using clustering




Clustering


John Snow’s Cholera Map

Clustering vs Hierarchical Clustering




Clustering Hierarchical Clustering

E

E1

E2

E3

E4

E7

E8




E



E1 E2


E7 E8

Agglomerative Clustering


1) Find Euclidean Distance between every node


2) Let X and Y be the nodes Closest to each other


3) Cluster X and Y store the distance


4) Repeat until every element is in one cluster



Time complexity θ(n^3)

Better Clustering than θ(n^3)?


Pre
-
process data


Canopy Clustering


Clusters faster


Improves Clusters


Faster Algorithms


BIRCH
(balanced iterative reducing and clustering using hierarchies)


CLARANS

(A Clustering Algorithm based on Randomized Search)

Real World Applications

•Google’s X laboratory


Artificial “Brain”


Deep Learning Algorithm


Identifies Cats based



on YouTube videos




Questions?

References

St. Clair, Daniel C. "Learning Programs: Teaching Computers to Acquire
Knowledge."
IEEE Potentials
. October (1992): 19
-
22. Print.


http://www.cse.chalmers.se/edu/year/2011/course/TDA231/


http://www.cs.unm.edu/~terran/downloads/classes/cs529
-
s11/papers/samuel_1959_B.pdf


http://www.werc.tu
-
darmstadt.de/fileadmin/user_upload/GROUP_WERC/LKE/tutorials/ML
-
tutorial
-
1
-
2.pdf


http://www.cs.princeton.edu/~schapire/talks/picasso
-
minicourse.pdf


http://www.cs.iastate.edu/~cs472/lectures/lecture05
-
game
-
2up.pdf



References (cont)

http://www.guardian.co.uk/news/datablog/2013/mar/15/john
-
snow
-
cholera
-
map



http://www.guardian.co.uk/news/datablog/interactive/2013/mar/15/cholera
-
map
-
john
-
snow
-
recreated

http://geoawesomeness.com/?p=3761


https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=
0CDoQFjAB&url=http%3A%2F%2Fwww.site.uottawa.ca%2F~nat%2FCours
es%2FCSI5387%2FML_Lecture_10.ppt&ei=MAmMUc6hF63QywGI54DgAg
&usg=AFQjCNF_DfVlmsnAkZX8WmXza9UisaoMdQ&bvm=bv.46340616,d.
aWc&cad=rja

http://research.microsoft.com/en
-
us/projects/languageunderstanding/