Programming for Social Scientists Lecture 7 UCLA Political Science 209-1: Programming for Social Scientists Winter 1999 Lars-Erik Cederman & Benedikt Stefansson

Programming for Social Scientists

Lecture 7UCLA Political Science 209-1: Programming for Social

Scientists

Winter 1999

Lars-Erik Cederman & Benedikt Stefansson

POL SCI 209-1 Cederman / Stefansson

2

Today's topics

• Network of agents

• 2D Space

• GUI: Zoom Rasters

Code example: GridIPD– Program that implements Cohen et al.'s

framework graphically


3

GridIPD: Adding 2D Spacemain

EZGraph

playerList

Tournament

ControlPanel

ObserverSwarm

Object2dDisplay

Raster

ColorMap

neighListDiscrete2d

ModelSwarm


4

Changes to GraphIPD...

• Spatially represented agents

• Single list with shuffling

• Network of neighbors

• Adaptation rather than evolution

• Raster display


5

Spatially represented agents

• Use these collection objects for players – Discrete2d

– List

• Assign each player a list of neighbors in 4 adjacent squares (i.e. Von Neuman neighborhood)

List

Discrete2d

List


6

Creating and collecting players

world = [Discrete2d createBegin: self];

[world setSizeX: worldSize Y: worldSize];

world = [world createEnd];

(HERE SKIP CODE TO CREATE PLAYERS AND PUT ON LIST)

[self shuffle: playerList];

index = [playerList begin: self];

for (x = 0; x < worldSize; x++)

for (y = 0; y < worldSize; y++) {

aPlayer = [index next];

[world putObject: aPlayer atX: x Y: y];

[aPlayer setX: x Y: y];

}

[index drop];

• Create grid object and set size

• Shuffle players• Put one player on

each space on grid


7

Using an index to iterate

index = [list begin: self];

while(obj=[index next])

(DO SOMETHING)

[index drop];

• Each List object can create index to iterate over collection

• Here [index next] moves to next member and returns member

• In while(…) when reach end of list index returns nil or false

• Drop index after use to reclaim memory


8

Shuffling agents on a list

-shuffle: list {

int j, k;

id temp;

j = [list getCount];

while (j>1) {

k = [uniformIntRand getIntegerWithMin: 0 withMax: j-1];

j--;

temp = [list atOffset: k];

[list atOffset: k put: [list atOffset: j]];

[list atOffset: j put: temp];

}

return self;

}


9

Shuffling agents on a list (cont)

60 1 2 3 4 5 7 8 j = [list getCount];j=9

k=5j=8

temp=

k=5j=8

j=8temp=

k = [...getIntegerWithMin: 0 withMax: j-1];j--




60 1 2 3 4 5 7 8

60 1 2 3 4 8 7 8

60 1 2 3 4 8 7 5

5

5

60 1 2 3 4 5 7 8


10

Shuffling agents - 2nd iteration

k=6j=7

temp=

k=5j=7

j=7temp=

k = [...getIntegerWithMin: 0 withMax: j-1];j--




60 1 2 3 4 8 7 5

70 1 2 3 4 8 7 5

70 1 2 3 4 8 6 5

6

6

60 1 2 3 4 8 7 5


11

Neighborhoods of agents

-setNeighborhoods {

id index, aPlayer;

index = [playerList begin: self];

while ((aPlayer=[index next])) {

[self setNeighborhood: aPlayer atDX: -1 DY: 0];

[self setNeighborhood: aPlayer atDX: 1 DY: 0];

[self setNeighborhood: aPlayer atDX: 0 DY: -1];

[self setNeighborhood: aPlayer atDX: 0 DY: 1];

}

[index drop];

return self;

}

• Each player has a list to hold neighbors

• Code here picks out players in “Von Neuman” neighborhood

• Note use of index


12

The Von Neuman neighborhood

NW N NE

EW

SW SE

is in location x,y

W

is in location x-1,y

N

is in location x,y-1

E

is in location x+1,y

Must correct for border agents

List

S


13

Populating neighborhoods

• ModelSwarm: Pick out neighbors, check if boundary conditions hold and call player to add to his list:

-setNeighborhood: (id) player atDX: (int) dx DY: (int) dy { int x, y;

x = [player getX];

y = [player getY];

if ([self validX: x+dx Y: y+dy]) {

[[player getNeighborhood] addLast: [world getObjectAtX:x+dx Y:y+dy]];

}

return self;

}

-(BOOL) validX: (int) x Y: (int) y {

return (((x >= 0) && (x < worldSize)) && ((y >= 0) && (y < worldSize)));

}


14

Adaptation: Player.m

-adaptType {

...

index = [neighbors begin: [self getZone]];

while ((neigh = [index next])) {

currentPayoff = [neigh getAveragePayoff];

if ((currentPayoff > bestPayoff)) {

bestPayoff = currentPayoff;

bestType = [neigh getPlayerType];

} else if (currentPayoff == bestPayoff)

if ([uniformDblRand getDoubleWithMin:0.0 withMax: 1.0]< 0.5){



}}

[index drop];

if (bestPayoff > [self getAveragePayoff])

newType = bestType;

return self;

}


15

Adaptation (cont)

From Player.m: -adaptType:while ((neigh = [index next])) {

currentPayoff = [neigh getAveragePayoff];

if ((currentPayoff > bestPayoff)) {



} else if (currentPayoff == bestPayoff) {

(BREAK TIES)

}

}

From Player.m: -updateType[self setPlayerType: newType]

• Player looks around neighborhood and picks neighbor type with highest payoff

• Put type value into temporary variable

• Finally set type to value of temporary variable


16

Adaptation (cont)

• To avoid creating new population use agents as “buffers”:– In -adaptType new type

is simply stored for future reference

• After all agents have found new type -updateType “creates” new population

type=2

type=11

type=2newType=12

type=13

= highest payoff


17

Objects to display raster image

Discrete2d

RasterHolds x,ylocations of agents

Object2dDisplay

Colormap

ModelSwarmObserverSwarm

Main purpose of Object2dDisplay to catch messages from Raster widget on a mouse-click. It then opens up a probe to the object that at this location.

Agent


18

The Space library classesSpace

Discrete2d

DblBuffer2d

Grid2d

ConwayLife Ca2d

Diffuse2d

Object2dDisplay

Value2dDisplay

Int2dFiler


19

Space

Discrete2d

DblBuffer2d

Grid2d

The basic 2d classes

• Discrete2d Provides basic 2d lattice

functions

• Subclasses– Grid2d

• Doesn’t allow you to overwrite members

– DblBuffer2d• Writes go to a buffer

• Then updates all at once


20

Some basic Discrete2d syntax

• setSizeX: i Y: j – Set the world size.

• getSizeX, getSizeY– Get the dimensions of

the lattice• fillWithValue: val, fillWithObject: obj– puts same value or object

at each site

• setDiscrete2d: o toFile: filename– Reads a PGM

formatted file and fills lattice with values

• copyDiscrete2d: a toDiscrete2d: b– Fills lattice b with

values from a


21

Creating a Colormap

colorMap = [Colormap create: self];

[colorMap setColor: 0 ToName: "red"];

[colorMap setColor: 1 ToName: "blue"];

[colorMap setColor: 2 ToName: "orange"];

[colorMap setColor: 3 ToName: "black"];

• Maps integer values used in simulation to color info used in display

• Can use– Color names

– RGB values

• Type is 0,1,2,3 so respective color appears

0 1 2 3

-drawSelfOn: (id) raster {[raster drawPointX: x Y: y Color: type];return self;}

From Player.m:


22

Creating Raster

worldRaster = [ZoomRaster create: self];

[worldRaster setColormap: colorMap];

[worldRaster setZoomFactor: zoomFactor];

[worldRaster setWidth: [modelSwarm getSize]

Height: [modelSwarm getSize]];

[worldRaster setWindowTitle: "Spatial PD"];

[worldRaster pack];

• Raster needs– Colormap

– Zoom factor (pixels per square)

– Dimensions

– Title


23

Creating Object2dDisplay

• Purpose: To catch mouse-clicks on raster and open probe to agent at this grid square

• Note that we need to define both Discrete2d, List, method to display agents and method to display probe

worldDisplay = [Object2dDisplay createBegin: self];

[worldDisplay setDisplayWidget: worldRaster];

[worldDisplay setDiscrete2dToDisplay: [modelSwarm getWorld]];

[worldDisplay setObjectCollection: [modelSwarm getPlayers]];

[worldDisplay setDisplayMessage: M(drawSelfOn:)];

worldDisplay = [worldDisplay createEnd];

[worldRaster setButton: ButtonRight Client: worldDisplay Message: M(makeProbeAtX:Y:)];


24

Exercise: Moore neighborhood

NW N NE

EW

SW SE

List

S

Now all 8 neighborsincluded

List can accomdatemore members

• In agent creation add code to define Moore neighborhood

• Hint: Consider for-next loop

Documents

Programming for Social Scientists Lecture 7 UCLA Political Science 209-1: Programming for Social Scientists Winter 1999 Lars-Erik Cederman & Benedikt Stefansson