Class Net

java.lang.Object
  |
  +--Net

public class Net

extends java.lang.Object

The Net class is in essence an n-dimensional array of Nodes.

Field Summary

int	cycleNum The number of cycles the algorithm has taken so far
int	dim The number of dimesions
Node[]	grid The actual array of Nodes
int	gridSize The number of cells (L^dim)
float	J The bond strength
int	L The side length of the Net
int	trueSize The number of cells including the virtual "+" and "-" cells (gridSize+2)

Constructor Summary

Net(java.io.File f, Wrapper didItFail)
Creates a Net from a previously saved file

Net(int size, float sigma, int d)
Creates a random Net

Net(int size, int d)
Creates an empty Net

Method Summary

void	clearFlows() "Resets" the Net by clearing all flows so it may be solved multiple times sequentially
java.lang.Object	clone() Creates an independent copy of the Net.
double	findEnergy() Finds the energy (aka max-flow value) of the Net once it has been solved
int[]	getCoords(Node node) Returns the coordinates of a given Node
Node	getNode(int[] coords) Returns a node given an array of its coordinates
ArrayHeap	globalUpdate() For Heaps
void	globalUpdate(Deque S) For Stacks
void	globalUpdate(Queue Q) For Queues
void	globalUpdate(Queue Q, boolean b) For double-sided queues.
static void	main(java.lang.String[] arg)
void	prStep(ArrayHeap H) For Heaps (uses gap)
void	prStep(ArrayHeap H, boolean gapFlag) For Heaps
void	prStep(Deque S) For Stacks
void	prStep(Queue Q) For Queues
void	prStep(Queue Q, boolean b) For double-sided queues
void	randNet(float cs) Changes the field strength of each Node in the Net by a small (or large) amount.
void	reTimes(float oldS, float newS) Gives a Net a new field strength by multilplying every Node's field strength b a conversion factor.
void	saveNet(java.io.File f) Writes the net to a file as the number of dimensions, followed by the side length, followed by the random field strengths, each on a new line
void	solve(int method, boolean gap, int updateFreq) The actual push-relabel algorithm
java.lang.String	toString() Not used in the actual applet, this method is included for use in debugging algorithmic changes.

Methods inherited from class java.lang.Object

equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

L

public int L

The side length of the Net

dim

public int dim

The number of dimesions

gridSize

public int gridSize

The number of cells (L^dim)

trueSize

public int trueSize

The number of cells including the virtual "+" and "-" cells (gridSize+2)

J

public float J

The bond strength

grid

public Node[] grid

The actual array of Nodes

cycleNum

public int cycleNum

The number of cycles the algorithm has taken so far

Constructor Detail

Net

public Net(java.io.File f,
           Wrapper didItFail)

Creates a Net from a previously saved file

Net

public Net(int size,
           float sigma,
           int d)

Creates a random Net

Parameters:

size - is the side length (L) of the Net

sigma - is the standard deviation of the Gaussian-distributed random field strengths

d - is the number of dimensions

Net

public Net(int size,
           int d)

Creates an empty Net

Parameters:

size - is the side length (L) of the Net

d - is the number of dimensions

Method Detail

randNet

public void randNet(float cs)

Changes the field strength of each Node in the Net by a small (or large) amount.

Parameters:

cs - is standard deviation of the Gaussian distribution governing the changes

clearFlows

public void clearFlows()

"Resets" the Net by clearing all flows so it may be solved multiple times sequentially

toString

public java.lang.String toString()

Not used in the actual applet, this method is included for use in debugging algorithmic changes. It shows the flow through every bond in a 2D Net, and a single face for a 3D.

Overrides:

toString in class java.lang.Object

solve

public void solve(int method,
                  boolean gap,
                  int updateFreq)

The actual push-relabel algorithm

Parameters:

method - the heuristic to use (heap, stack, queue, etc.)

gap - whether or not to use the gap (only meaningful if a heap-based heuristic is used)

updateFreq - the number of push-relabel cycles to elapse between global updates (-1 gives no global updates)

prStep

public void prStep(ArrayHeap H)

For Heaps (uses gap)

prStep

public void prStep(ArrayHeap H,
                   boolean gapFlag)

For Heaps

Parameters:

gapFlag - specifies whether the gap is to be used or not

globalUpdate

public ArrayHeap globalUpdate()

For Heaps

prStep

public void prStep(Deque S)

For Stacks

globalUpdate

public void globalUpdate(Deque S)

For Stacks

prStep

public void prStep(Queue Q)

For Queues

globalUpdate

public void globalUpdate(Queue Q)

For Queues

prStep

public void prStep(Queue Q,
                   boolean b)

For double-sided queues

globalUpdate

public void globalUpdate(Queue Q,
                         boolean b)

For double-sided queues. Note that this global update takes twice as long as the others, since the positive and negative excesses have different height fields.

reTimes

public void reTimes(float oldS,
                    float newS)

Gives a Net a new field strength by multilplying every Node's field strength b a conversion factor.

getNode

public Node getNode(int[] coords)

Returns a node given an array of its coordinates

getCoords

public int[] getCoords(Node node)

Returns the coordinates of a given Node

findEnergy

public double findEnergy()

Finds the energy (aka max-flow value) of the Net once it has been solved

clone

public java.lang.Object clone()

Creates an independent copy of the Net. Note that flows are *not* copied over, so this should only be used before any algorithm steps have occurred.

Overrides:

clone in class java.lang.Object

saveNet

public void saveNet(java.io.File f)

Writes the net to a file as the number of dimensions, followed by the side length, followed by the random field strengths, each on a new line

main

public static void main(java.lang.String[] arg)