linalg:
PortedBLAS
spr()
linalg:
PortedMatrix
copy()
andtoString()
, and implementedclone()
to make matrices usable within RDD functions as a zero value parameterlinalg:
PortedRowMatrix
computeCovariance()
,computeGramianMatrix()
andtriuToFull()
util:
New utility equivalent toscala.collection.Iterator
fill()
stat:
Portedcorr()
andPearsonCorrelation
class
linalg:
Use native Lua 1-basedcolPtrs
indexes inSparseMatrix
- #56 Java array
binarySearch()
claimed to use 1-based index params and an exclusive end-index, but used 0-based index params and an inclusive end-index
linalg:
DenseVector
andSparseVector
are now clonable, making them usable as a zero value in RDD functions; required byRowMatrix
- Document how to train a K-means model within Redis at examples/redis-kmeans
- Document how to run test suites within Redis at spec-redis
- #43 Support loading LIBSVM files from
MLUtils
DenseMatrix
andSparseMatrix
map() and update() supportDenseMatrix
diag() and eye() support
- #47 Cannot train a K-means model within Redis due to unnecessary os.date() dependency (Redis interop)
- #31
[linalg:]
SupportDenseMatrix
,Matrices
,Matrix
, andSparseMatrix
modules - [regression:] Support LabeledPoint class
- DenseVector and SparseVector toString() format changed to conform to Apache Spark (instead of Lua conventions), now that LabeledPoint unit test coverage requires compatibility
- Upgraded to Stuart 2.0.0
- #38 Vectors.dense(...) with varargs error: no table.pack() function (eLua interop)
- Upgrade to Stuart 1.0.0 with changes to class framework
- #28 Train a K-means model
- #34 Remove all use of module upvalues, so that modules can be more easily transpiled to C then burned into ROM and chipsets (eLua LTR interop)
- #23 K-means cluster centers load in a non-deterministic sort order due to missing sort
- #25 VectorWithNorm initializer fails to compute norm when not explicitly provided
- Lua 5.3 support
- #17
stat:
PortedMultivariateOnlineSummarizer
class,RowMatrix
numCols(), numRows(), computeColumnSummaryStatistics(), and updateNumRows(), andstatistics
colStats() - #20 Package and deploy releases to npmjs.com and jsDelivr
- #15 Remove stuart-sql LuaRocks dependency. It is still used when present, but no longer required.
- Upgrade to Stuart 0.1.7
clustering:
PortedKMeansModel
util:
PortedLoader
, which can load aKMeansModel
from a Parquet file or directory of files, whether local or WebHDFSutil:
PortedNumericParser
util:
PortedStringTokenizer
clustering:
PortedKMeans
fastSquaredDistance()
,findClosest()
andpointCost()
linalg:
PortedBLAS
axpy()
,dot()
, andscal()
linalg:
PortedVectors
sqdist()
linalg:
ImplementVectorWithNorm
equality (__eq
)util:
PortedMLUtils
EPSILON
andfastSquaredDistance()
clustering:
Fixed KMeans getters and setters to match RDD API, not DataFrame APIlinalg:
Vector types are now 0-based like Apache Spark- Consolidate Apache Spark and Stuart unit tests into a single unified folder hierarchy
linalg:
VectornumActives
andnumNonzeros
fields were not updating after changes to vector.BLAS
functions mutate vectors. Changed fields to functions so that they're computed each time.
clustering:
PortedVectorWithNorm
datatypelinalg:
PortedDenseVector
andSparseVector
datatypes, plus theirVectors
factory
(formatted per keepachangelog-1.1.0)