public class DistributedLDAModel extends LDAModel
Distributed model fitted by LDA.
This type of model is currently only produced by Expectation-Maximization (EM).
This model stores the inferred topics, the full training dataset, and the topic distribution for each training document.
param: oldLocalModelOption Used to implement oldLocalModel as a lazy val, but keeping
copy() cheap.
| Modifier and Type | Method and Description |
|---|---|
IntParam |
checkpointInterval()
Param for set checkpoint interval (>= 1) or disable checkpoint (-1).
|
DistributedLDAModel |
copy(ParamMap extra)
Creates a copy of this instance with the same UID and some extra params.
|
DoubleArrayParam |
docConcentration()
Concentration parameter (commonly named "alpha") for the prior placed on documents'
distributions over topics ("theta").
|
Param<java.lang.String> |
featuresCol()
Param for features column name.
|
int |
getCheckpointInterval() |
double[] |
getDocConcentration() |
java.lang.String |
getFeaturesCol() |
int |
getK() |
double |
getLearningDecay() |
double |
getLearningOffset() |
int |
getMaxIter() |
protected LDAModel |
getModel()
Returns underlying spark.mllib model, which may be local or distributed
|
Vector |
getOldDocConcentration()
Get docConcentration used by spark.mllib LDA
|
LDAOptimizer |
getOldOptimizer() |
double |
getOldTopicConcentration()
Get topicConcentration used by spark.mllib LDA
|
boolean |
getOptimizeDocConcentration() |
java.lang.String |
getOptimizer() |
long |
getSeed() |
double |
getSubsamplingRate() |
double |
getTopicConcentration() |
java.lang.String |
getTopicDistributionCol() |
boolean |
isDistributed()
Indicates whether this instance is of type
DistributedLDAModel |
IntParam |
k()
Param for the number of topics (clusters) to infer.
|
DoubleParam |
learningDecay()
Learning rate, set as an exponential decay rate.
|
DoubleParam |
learningOffset()
A (positive) learning parameter that downweights early iterations.
|
static DistributedLDAModel |
load(java.lang.String path) |
double |
logPrior()
Log probability of the current parameter estimate:
log P(topics, topic distributions for docs | Dirichlet hyperparameters)
|
IntParam |
maxIter()
Param for maximum number of iterations (>= 0).
|
protected LocalLDAModel |
oldLocalModel()
Underlying spark.mllib model.
|
BooleanParam |
optimizeDocConcentration()
Indicates whether the docConcentration (Dirichlet parameter for
document-topic distribution) will be optimized during training.
|
Param<java.lang.String> |
optimizer()
Optimizer or inference algorithm used to estimate the LDA model.
|
static MLReader<DistributedLDAModel> |
read() |
LongParam |
seed()
Param for random seed.
|
DoubleParam |
subsamplingRate()
Fraction of the corpus to be sampled and used in each iteration of mini-batch gradient descent,
in range (0, 1].
|
java.lang.String[] |
supportedOptimizers()
Supported values for Param
optimizer. |
LocalLDAModel |
toLocal()
Convert this distributed model to a local representation.
|
DoubleParam |
topicConcentration()
Concentration parameter (commonly named "beta" or "eta") for the prior placed on topics'
distributions over terms.
|
Param<java.lang.String> |
topicDistributionCol()
Output column with estimates of the topic mixture distribution for each document (often called
"theta" in the literature).
|
double |
trainingLogLikelihood()
Log likelihood of the observed tokens in the training set,
given the current parameter estimates:
log P(docs | topics, topic distributions for docs, Dirichlet hyperparameters)
|
StructType |
validateAndTransformSchema(StructType schema)
Validates and transforms the input schema.
|
void |
validateParams() |
MLWriter |
write()
Returns an
MLWriter instance for this ML instance. |
describeTopics, describeTopics, estimatedDocConcentration, logLikelihood, logPerplexity, setFeaturesCol, setSeed, sqlContext, topicsMatrix, transform, transformSchema, uid, vocabSizetransform, transform, transformtransformSchemaclone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, waitclear, copyValues, defaultCopy, defaultParamMap, explainParam, explainParams, extractParamMap, extractParamMap, get, getDefault, getOrDefault, getParam, hasDefault, hasParam, isDefined, isSet, paramMap, params, set, set, set, setDefault, setDefault, shouldOwntoStringinitializeIfNecessary, initializeLogging, isTraceEnabled, log_, log, logDebug, logDebug, logError, logError, logInfo, logInfo, logName, logTrace, logTrace, logWarning, logWarningsavepublic static MLReader<DistributedLDAModel> read()
public static DistributedLDAModel load(java.lang.String path)
protected LocalLDAModel oldLocalModel()
LDAModeloldLocalModel in class LDAModelprotected LDAModel getModel()
LDAModelpublic LocalLDAModel toLocal()
WARNING: This involves collecting a large topicsMatrix to the driver.
public DistributedLDAModel copy(ParamMap extra)
Paramspublic boolean isDistributed()
LDAModelDistributedLDAModelisDistributed in class LDAModelpublic double trainingLogLikelihood()
Notes:
- This excludes the prior; for that, use logPrior.
- Even with logPrior, this is NOT the same as the data log likelihood given the
hyperparameters.
- This is computed from the topic distributions computed during training. If you call
logLikelihood() on the same training dataset, the topic distributions will be computed
again, possibly giving different results.
public double logPrior()
public MLWriter write()
MLWritableMLWriter instance for this ML instance.public IntParam k()
public int getK()
public DoubleArrayParam docConcentration()
This is the parameter to a Dirichlet distribution, where larger values mean more smoothing (more regularization).
If not set by the user, then docConcentration is set automatically. If set to
singleton vector [alpha], then alpha is replicated to a vector of length k in fitting.
Otherwise, the docConcentration vector must be length k.
(default = automatic)
Optimizer-specific parameter settings:
- EM
- Currently only supports symmetric distributions, so all values in the vector should be
the same.
- Values should be > 1.0
- default = uniformly (50 / k) + 1, where 50/k is common in LDA libraries and +1 follows
from Asuncion et al. (2009), who recommend a +1 adjustment for EM.
- Online
- Values should be >= 0
- default = uniformly (1.0 / k), following the implementation from
https://github.com/Blei-Lab/onlineldavb.
public double[] getDocConcentration()
public Vector getOldDocConcentration()
public DoubleParam topicConcentration()
This is the parameter to a symmetric Dirichlet distribution.
Note: The topics' distributions over terms are called "beta" in the original LDA paper by Blei et al., but are called "phi" in many later papers such as Asuncion et al., 2009.
If not set by the user, then topicConcentration is set automatically. (default = automatic)
Optimizer-specific parameter settings:
- EM
- Value should be > 1.0
- default = 0.1 + 1, where 0.1 gives a small amount of smoothing and +1 follows
Asuncion et al. (2009), who recommend a +1 adjustment for EM.
- Online
- Value should be >= 0
- default = (1.0 / k), following the implementation from
https://github.com/Blei-Lab/onlineldavb.
public double getTopicConcentration()
public double getOldTopicConcentration()
public java.lang.String[] supportedOptimizers()
optimizer.public Param<java.lang.String> optimizer()
For details, see the following papers:
- Online LDA:
Hoffman, Blei and Bach. "Online Learning for Latent Dirichlet Allocation."
Neural Information Processing Systems, 2010.
http://www.cs.columbia.edu/~blei/papers/HoffmanBleiBach2010b.pdf
- EM:
Asuncion et al. "On Smoothing and Inference for Topic Models."
Uncertainty in Artificial Intelligence, 2009.
http://arxiv.org/pdf/1205.2662.pdf
public java.lang.String getOptimizer()
public Param<java.lang.String> topicDistributionCol()
This uses a variational approximation following Hoffman et al. (2010), where the approximate distribution is called "gamma." Technically, this method returns this approximation "gamma" for each document.
public java.lang.String getTopicDistributionCol()
public DoubleParam learningOffset()
public double getLearningOffset()
public DoubleParam learningDecay()
public double getLearningDecay()
public DoubleParam subsamplingRate()
Note that this should be adjusted in synch with LDA.maxIter
so the entire corpus is used. Specifically, set both so that
maxIterations * miniBatchFraction >= 1.
Note: This is the same as the miniBatchFraction parameter in
OnlineLDAOptimizer.
Default: 0.05, i.e., 5% of total documents.
public double getSubsamplingRate()
public BooleanParam optimizeDocConcentration()
public boolean getOptimizeDocConcentration()
public StructType validateAndTransformSchema(StructType schema)
schema - input schemapublic void validateParams()
validateParams in interface Paramspublic LDAOptimizer getOldOptimizer()
public Param<java.lang.String> featuresCol()
public java.lang.String getFeaturesCol()
public IntParam maxIter()
public int getMaxIter()
public LongParam seed()
public long getSeed()
public IntParam checkpointInterval()
public int getCheckpointInterval()