Frontiers in Mathematics and Computer Science

unknownlippsAI and Robotics

Oct 16, 2013 (3 years and 11 months ago)

80 views

Frontiers in Mathematics and
Computer Science

Salt Lake City Public Library, SLC, Utah


Nazmus

Saquib

Scientific Computing and Imaging Institute


welcome back!


t
oday we will



experiment with some code



l
earn a bit about graph theory and genetic
algorithm



d
iscuss the implications of mathematics
research

i
nstalling python and
pygame




http
://www.python.org/download
/





http://
www.pygame.org/download.shtml



python is a programming language



suitable for beginning and learning programming



w
e will play with some python examples today

a
genda


day 2


m
athematics


c
haos theory


b
utterfly effect


w
eather forecast


f
ractal music


L
-
systems


s
ocial interactions (in
facebook
)



g
raph theory


s
ocial interactions example (continued)




agenda


day 2


c
omputer science


machine learning


b
ig data


g
enetic algorithms



d
ata mining


s
entiment analysis


digital humanities



graph theory


i
n the context of social interactions



c
an we predict the behavior of a group of
people? (given some information)



g
roup dynamics



graph network



jargon


n
ode

and
edge

http://pc57724.uni
-
regensburg.de/morgan/teaching/CS104
-
Social_Networking.pdf

c
ulture hubs


d
egree

of a
node

http://en.wikipedia.org/wiki/File:Scale
-
free_network_sample.png

(very primary
)
types of analysis


power


(who’s The Guy?!)


r
elated to the
degree

of a graph



c
loseness


h
ow many people do I need to know to reach
someone else
asap
?





http://pc57724.uni
-
regensburg.de/morgan/teaching/CS104
-
Social_Networking.pdf

(primary)
types of analysis


betweeness


who can get me to the most important people
asap
?



asap
:

shortest path

in the graph



number of times I need to go through someone to
reach someone else


(primary)
types of analysis


betweeness


(only equation in the slides, I promise!)

this is to show you how easy it is to calculate such metrics

e
xample


15
th

century Florence


Medici family was less powerful than others



t
hey ended up dominating



w
hy is that so?



b
etweeness

score



Medici: 0.52



second largest: 0.25



moral: networking is important!



Medici held the network together

that finishes our math portion




artificial intelligence


machine learning

is the development of
algorithms from which programs can
learn



what can they learn?



what can they do with the training?



t
raining datasets

invitation to
big data


we deal with
exabytes

of data nowadays




1 exabyte
= 1

099

511

627

776
megabytes




2147483 hard disks (that are each 500 GB) !!





h
ow do we make sense of such a huge amount of information?




opportunities in
supercomputing

and
machine learning

flavor of artificial intelligence


Terminator 2 was not quite right, robots
haven’t taken over yet



but we can use AI in other ways



evolutionary algorithms



set a goal, evolve your given information
towards the goal



g
enetic algorithm

genetic algorithm


s
ay, you would like to break someone’s
password



you can try all random combinations



or you can do some
intelligent guesses



h
ow can we simulate this process for a
computer?

simple genetic algorithm


start
with “;
wql
*
opqlq




end goal: “hello world”

;

w

q

l

*

o

p

q

l

q

h

e

l

l

o

w

o

r

l

d

genetic algorithm


treat these characters as genes!



genes can mutate, right?

;

w

q

l

*

o

p

q

l

q

;

w

q

l

*

o

o

q

l

q

genetic algorithm


but wait, the program should not accept
every mutation



h
ow does it know it is closer to the goal?



h
ow can we find the difference between
two sets?



Euclidean distance


genetic algorithm


fitness test: is a gene fit to pass?



If the difference between source and target is
lower, we accept the mutation.



intermediate results are important too!



i
n reality, you would derive a good fitness function
that would produce “intelligent” results



if you were writing a password breaker, you
wouldn’t know the password to begin with!

genetic algorithm


t
ext evolution example (textevolve.py)



m
usic evolution example
(music_evolve.py)

research in mathematics


discussion

end of day 2


resources can be found at


nsaquib.com/presentations


code examples


things to try out



thanks for attending!