|
||||||||||
PREV NEXT | FRAMES NO FRAMES |
Factorizer
s, provides ID to index mappingOnlineLogisticRegression
learners.AdaptiveLogisticRegression.DEFAULT_THREAD_COUNT
and AdaptiveLogisticRegression.DEFAULT_POOL_SIZE
containsKey
method on the map returned by AbstractJob.parseArguments(String[])
;
AbstractJob.parseArguments(String[])
is called.
AbstractJob.parseArguments(String[])
is called.
AbstractJob.parseArguments(String[])
is called.
AbstractJob.parseArguments(String[])
is called.
ItemSimilarity.allSimilarItemIDs(long)
as candidate itemsRecommenderEvaluator
which computes the average absolute
difference between predicted and actual ratings for users.BooleanUserPreferenceArray
but stores preferences for one item (all item IDs the same) rather
than one user.GenericUserPreferenceArray
but stores, conceptually, BooleanPreference
objects which
have no associated preference value.DataModel
implementation to be used in an evaluation, given training data.
Recommender
implementation to be evaluated, using the given DataModel
.
Collection
if the method parameter is null
.
Retriever
.
Retriever
and with given maximum size.
ModelTrainer
with two TopicModel
instances:
one from the previous iteration, the other empty.CachingCVB0PerplexityMapper
, to aid in debugging.ItemSimilarity
implementation.ItemSimilarity
.
ItemSimilarity
.
Recommender
which caches the results from another Recommender
in memory.UserNeighborhood
implementation.UserSimilarity
implementation.UserSimilarity
.
UserSimilarity
.
SequenceFile
format.
GenericRecommenderIRStatsEvaluator.evaluate(RecommenderBuilder, DataModelBuilder, DataModel, IDRescorer, int, double, double)
to
have it attempt to compute a reasonable threshold.
n-1
scores, where
n
is equal to numCategories()
, given an input
vector instance
.
n-1
, for each row of a matrix, where n
is equal
to numCategories()
.
n
scores, where
n
is numCategories()
, given an input vector
instance
.
n
scores, where
n
is numCategories()
, given an input vector
instance
.
n
probabilities, one for each category.
AbstractVectorClassifier.classify(Vector)
would return a vector with only one element.
closeables
(to prevent repeating close attempts), re-throw the
last one at the end.
Parametered.createParameters(String,org.apache.hadoop.conf.Configuration)
on parameter parmetered, and then recur down its composite tree to invoke
Parametered.createParameters(String,org.apache.hadoop.conf.Configuration)
and Parametered.configure(org.apache.hadoop.conf.Configuration)
on
each composite part.
Reducer
classReducer
for PFPGrowth which updates the status as well as writes the
patterns generated by the algorithmPair
s whose second element is a count.to-1
.SequenceFileDirValueIterator
Version
.
Version
.
SequenceFile
format.
CachingCVB0Mapper
for more details on scalability and room for improvement.Preference
s
for items.DataModel
to be
used while evaluating a Recommender
.MatrixWritable
.StringTuple
s.The
SequenceFile
input should have a Text
key
containing the unique document identifier and a
Text
value containing the whole document.SequenceFile
, select k vectors and write them to the
output file as a Kluster
representing the initial centroid to use.WritableComparable
encapsulating two items.Writable
encapsulating an item ID and a preference value.Recommender
's recommendations.
FastByIDMap
with default capacity.
FastByIDMap
whose capacity can accommodate the given number of entries without rehash.
FastIDSet
with default capacity.
Map
implementation, based on algorithms described in Knuth's "Art of Computer
Programming", Vol.FastMap
with default capacity.
DataModel
backed by a delimited file.IDMigrator
backed by a file.ItemSimilarity
backed by a comma-delimited file.FileLineIterable
over a given file, assuming a UTF-8 encoding.
FileLineIterable
over a given file, assuming a UTF-8 encoding.
FileLineIterable
over a given file, using the given encoding.
FileLineIterator
over a given file, assuming a UTF-8 encoding.
FileLineIterator
over a given file, assuming a UTF-8 encoding.
FileLineIterator
over a given file, using the given encoding.
FileItemSimilarity
FPGrowth
algorithmObject
for each level of the recursive
FPGrowth
algorithm to reduce allocation overhead.FullRunningAverage
to add a running standard deviation computation.DictionaryVectorizer
job
DataModel
which uses given user data as its data source.GenericDataModel
from the given users (and their preferences).
GenericDataModel
from the given users (and their preferences).
GenericBooleanPrefDataModel.toDataMap(DataModel)
with GenericBooleanPrefDataModel.GenericBooleanPrefDataModel(FastByIDMap)
GenericItemBasedRecommender
which is appropriate for use when no notion of preference
value exists in the data.GenericUserBasedRecommender
which is appropriate for use when no notion of preference
value exists in the data.DataModel
which uses a given List
of users as its data source.GenericDataModel
from the given users (and their preferences).
GenericDataModel
from the given users (and their preferences).
GenericDataModel.toDataMap(DataModel)
with GenericDataModel.GenericDataModel(FastByIDMap)
Recommender
which uses a given
DataModel
and
ItemSimilarity
to produce recommendations.GenericUserPreferenceArray
but stores preferences for one item (all item IDs the same) rather
than one user.GenericItemSimilarity.ItemItemSimilarity
which takes a static list of precomputed item similarities and bases its
responses on that alone.GenericItemSimilarity
from a precomputed list of GenericItemSimilarity.ItemItemSimilarity
s.
GenericItemSimilarity.GenericItemSimilarity(Iterable)
, but will only keep the specified number of similarities
from the given Iterable
of similarities.
GenericItemSimilarity.ItemItemSimilarity
implementation and a
DataModel
, rather than a list of GenericItemSimilarity.ItemItemSimilarity
s.
GenericItemSimilarity.GenericItemSimilarity(ItemSimilarity, DataModel)
)}, but will only keep the specified
number of similarities from the given DataModel
.
Preference
encapsulating an item and preference value.RecommendedItem
.n
preferences, then evaluate the IR
statistics based on a DataModel
that does not have these values.Recommender
which uses a given DataModel
and UserNeighborhood
to produce recommendations.GenericItemPreferenceArray
but stores preferences for one user (all user IDs the same) rather
than one item.AbstractJob.parseArguments(String[])
.
DataModel.getMaxPreference()
DataModel.getMinPreference()
AbstractJob.parseArguments(String[])
.
Rescorer
which operates on long
primitive IDs, rather than arbitrary Object
s.CVB0Driver
, but sequentially, in memory.AbstractJob.parseArguments(String[])
FastIDSet
.
WritableComparable
which encapsulates an ordered pair of signed integers.Closeable
too,
where file is wiped on close and thus the disk resource is released
('closed').Recommender
's recommendations.true
to exclude the given thing.
true
to exclude the given thing.
AggregateAndRecommendReducer
automatically exclude themItemSimilarity.itemSimilarity(long, long)
.
ItemAverageRecommender
, except that estimated preferences are adjusted for the users' average
preference value.DistributedCache
.
LongPrimitiveArrayIterator
over an entire array.
long
primitives in the style of an Iterator
-- as
opposed to iterating over Long
.DataSource
by name from JNDI.
Refreshable
to the given collection of Refreshable
s if it is not
already there and immediately refreshes it.
RandomAccessSparseVector
s into the complete Document
RandomAccessSparseVector
TopicModel
and use it to iteratively learn the p(topic|term, doc)
distribution for documents (this can be done in parallel across many documents, as the
"read-only" model is, well, read-only.SharingMapper
s.len
lowest bytes
of val
.
RecommendedItem
PersistenceStrategy
which does nothing.Rescorer
which always returns the original score.AbstractJob.parseArguments(String[])
addOption
methods.
PathFilter
.SequenceFileDirIterable
and the like to select whether the input path specifies a
directory to list, or a glob pattern.Factorization
sPlusAnonymousUserDataModel
which allow multiple concurrent anonymous requests.DataModel
decorator class is useful in a situation where you wish to recommend to a user that
doesn't really exist yet in your actual DataModel
.Preference
encapsulates an item and a preference value, which indicates the strength of the
preference for it.Preference
.GenericOptionsParser
.
FastByIDMap
data structure which maps user IDs
to preferences.
SequenceFile
format.
ResultSet
, Statement
and Connection
(if not null) and logs (but does not
rethrow) any resulting SQLException
.
SequenceFile
, randomly select k vectors and
write them to the output file as a Kluster
representing the
initial centroid to use.Recommender.recommend(long, int, org.apache.mahout.cf.taste.recommender.IDRescorer)
, with a
Rescorer
that does nothing.
Writable
which encapsulates a list of RecommendedItem
s.Recommender
to be
evaluated based on the given DataModel
.Recommender
's recommendations.Recommender
's performance, including precision, recall and
f-measure.Refreshable.refresh(java.util.Collection)
and is the entire body of
that method.
Refreshable
.RecommenderIRStatsEvaluator
.FileDataModel.setPreference(long, long, float)
.
DataModel.removePreference(long, long)
(Object, Object)}.
Rescorer
simply assigns a new "score" to a thing like an ID of an item or user which a
Recommender
is considering returning as a top recommendation.RecommenderEvaluator
which computes the "root mean squared"
difference between predicted and actual ratings for users.ClusterClassifier
to classify input vectors into their
respective clusters.
RunningAverage
by adding standard deviation too.SamplingCandidateItemsStrategy.NO_LIMIT_FACTOR
) for all factors, except
candidatesPerUserFactor
which defaults to SamplingCandidateItemsStrategy.DEFAULT_FACTOR
.
Iterable
whose Iterable.iterator()
returns only some subset of the elements that
it would, as determined by a iterator rate parameter.Iterator
and returns only some subset of the elements that it would, as determined by a
iterator rate parameter.LongPrimitiveIterator
and returns only some subset of the elements that it would,
as determined by a sampling rate parameter.Iterable
counterpart to SequenceFileDirIterator
.SequenceFileIterator
, but iterates not just over one sequence file, but many.FileSystem.listStatus(Path)
or
FileSystem.globStatus(Path)
to obtain list of files to iterate over
(depending on pathType parameter).
Iterable
counterpart to SequenceFileDirValueIterator
.SequenceFileValueIterator
, but iterates not just over one
sequence file, but many.FileSystem.listStatus(Path)
or
FileSystem.globStatus(Path)
to obtain list of files to iterate over
(depending on pathType parameter).
Iterable
counterpart to SequenceFileIterator
.SequenceFileIterable.SequenceFileIterable(Path, boolean, Configuration)
but key and value instances are not reused
by default.
Iterator
over a SequenceFile
's keys and values, as a Pair
containing key and value.Writable
key and Writable
value, and writes them into a SequenceFile
Iterable
counterpart to SequenceFileValueIterator
.SequenceFileValueIterable.SequenceFileValueIterable(Path, boolean, Configuration)
but instances are not reused
by default.
Iterator
over a SequenceFile
's values only.Preference
FileDataModel
maintains; it does not modify any data on disk.
DataModel.setPreference(long, long, float)
.
PreferenceInferrer
to the UserSimilarity
implementation.
MultithreadedSharingMapper
.BatchItemSimilarities
implementationVectorOrPrefWritable
actually a column from that matrix has to be used but as the similarity matrix is symmetric,
we can use a row instead of having to transpose itFeatureVectorEncoder
the
input and writes it to the output as a sequence file.Iterator
.
Iterator.next()
repeatedly.PearsonCorrelationSimilarity
, but compares relative ranking of preference values instead of
preference values themselves.EuclideanDistanceMeasure
but it does not take the square root.MemoryUtil.startMemoryLogger(long)
or
MemoryUtil.startMemoryLogger()
.
Recommender
that uses matrix factorization (a projection of users
and items onto a feature space)AbstractJob.parseArguments(String[])
Weight
based on term frequency onlyList
s for arrays in Map
values .
StringTuple
The input documents has to be
in the SequenceFile
format
Matrix
of counts of occurrences of (topic, term) pairs.FrequentPatternMaxHeap
FPTree
This reduces plenty of space and speeds up
Map/Reduce of PFPGrowth
algorithm by reducing data size passed from the Mapper to the reducer where
FPGrowth
mining is doneVectorDistanceMapper
, except it outputs
<input, Vector>, where the vector is a dense vector contain one entry for every seed vector
|
||||||||||
PREV NEXT | FRAMES NO FRAMES |