org.knowceans.dirichlet.lda
Class LdaMarkovState

java.lang.Object
  org.knowceans.dirichlet.lda.LdaMarkovState

All Implemented Interfaces:

java.io.Serializable

Direct Known Subclasses:

AtmMarkovState, LdaMarkovStateHyper

public class LdaMarkovState
extends java.lang.Object
implements java.io.Serializable

LdaMarkovState represents the state of the gibbs sampler of the lda process used for estimating unknown documents (word vectors).

Author:

gregor

See Also:

Serialized Form

Field Summary

int[][]	nd n_m,k: nd[m][k] number of words (not terms!)
int[]	ndsum n_m: ndsum[m] total number of words in document m.
int[][]	nw n_k,t: nw[k][t] number of instances of term t assigned to topic k.
int[]	nwsum n_k: nwsum[k] total number of words (not terms!)
private static long	serialVersionUID
int	V V: size of vocabulary.
int[][]	w w_m,n: word vectors of the corpus.
int[][]	z z_m,n: topic assignments z_m,n for each word (= term occurrence).

Constructor Summary

LdaMarkovState()
Initialise markov state for loading or setting by parameters (subclasses only).

LdaMarkovState(int[][] w, int V, int[][] z, int K)
Initialise the object from the data given.

LdaMarkovState(LdaMarkovState state)
Copy constructor that copies all internal fields from state to this.

LdaMarkovState(java.lang.String ldaBase)
Set up the markov chain from a file.

Method Summary

void	copyTo(LdaMarkovState cps) Copy the information contained in this to the fields of the argument object.
void	init(int K, java.util.Random rand) Random allocation and initialisaton of the state count vectors.
protected void	initNd(int M, int K, java.util.Random rand)
protected void	initNw(int M, int K)
void	load(java.lang.String filename) Load state arrays from file.
void	recalculate(int[][] z) Recalculates the topic statistics from the word-topic associations.
void	save(java.lang.String filename) Save (optionally compressed) file.

Methods inherited from class java.lang.Object

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

Field Detail

serialVersionUID

private static final long serialVersionUID

See Also:

Constant Field Values

w

public int[][] w

w_m,n: word vectors of the corpus.

V

public int V

V: size of vocabulary.

z

public int[][] z

z_m,n: topic assignments z_m,n for each word (= term occurrence).

nw

public int[][] nw

n_k,t: nw[k][t] number of instances of term t assigned to topic k. In subclasses, this can be generalised as the number of associations between topics and latent-semantic "minor" items (words, links, recommendations).

nd

public int[][] nd

n_m,k: nd[m][k] number of words (not terms!) in document m assigned to topic k. This should be equal to the document length. In subclasses, this can be generalised as the number associations between latent-semantic "major" items (documents, authors, searchers, recommenders) and topics.

nwsum

public int[] nwsum

n_k: nwsum[k] total number of words (not terms!) assigned to topic k.

ndsum

public int[] ndsum

n_m: ndsum[m] total number of words in document m.

Constructor Detail

LdaMarkovState

public LdaMarkovState()

Initialise markov state for loading or setting by parameters (subclasses only).

LdaMarkovState

public LdaMarkovState(java.lang.String ldaBase)

Set up the markov chain from a file.

Parameters:

ldaBase -

LdaMarkovState

public LdaMarkovState(LdaMarkovState state)

Copy constructor that copies all internal fields from state to this.

Parameters:

state -

LdaMarkovState

public LdaMarkovState(int[][] w,
                      int V,
                      int[][] z,
                      int K)

Initialise the object from the data given.

Parameters:

w -

V -

z -

K -

Method Detail

recalculate

public void recalculate(int[][] z)

Recalculates the topic statistics from the word-topic associations.

Parameters:

z -

copyTo

public void copyTo(LdaMarkovState cps)

Copy the information contained in this to the fields of the argument object.

Parameters:

cps -

load

public void load(java.lang.String filename)

Load state arrays from file.

Parameters:

file -

save

public void save(java.lang.String filename)

Save (optionally compressed) file.

Parameters:

file -

init

public void init(int K,
                 java.util.Random rand)

Random allocation and initialisaton of the state count vectors.

Parameters:

K -

rand -

hasNw -

initNd

protected void initNd(int M,
                      int K,
                      java.util.Random rand)

initNw

protected void initNw(int M,
                      int K)