Export - CSV (All fields)
Export - CSV (Current fields)
SYSTEMDS-1827
Automatically determine max memory/compute bandwidthSYSTEMDS-1558
Order statistics over compressed matrix blocks
Cache-conscious post processing tsmmSYSTEMDS-1546
Parfor result partitioning rewrite not applied on spark
Incorrect leftindexing memory estimate w/ empty input in loops
Parfor remote result merge fails w/ stackoverflowerror from rdd union
Performance compressed rowsums
Performance compressed tsmm w/ ddc column groupsSYSTEMDS-1529
Accuracy score sometimes lower than expected in test_svm of test_mllearn_numpy.py
Missing codegen plan cleanup for multi-aggregate templates
Wrong size propagation of row and multi-agg codegen templates
Corrupted sparse matrix representations
Corrupted input file names in old and new mlcontext apis
Improve temporary memory handling CLA MV and VM operations
Improve output size calculation in conv2d & max_pool2d
Generalize cell template for sideways row vectors inputs
Wrong cost estimation of fused operation DAGs
Create additional common unary and binary row vector operations
Performance spark rowwise codegen instructions
Add Pygments license in source distribution
Support log and exp row vector operations
Additional meta operator template: SpoofRowwise
Support cross-partition multi-aggregates with partial shared reads
Create ParserFactory
Allow source imports for jar scripts
Incorrect codegen cost model (number and size of inputs)
Index out of bounds error in test_naive_bayes1 of test_mllearn_numpy.py
Wrong final aggregation on multi-threaded cellwise codegen templates
Avoid dense allocation of empty sideway inputsSYSTEMDS-1468
Add new 1D/2D "Scale & Shift" layers
optimizer changes the output of table funciton
Invalid frame textcell write/read for tokens w/ spaces
Rename `batch_norm.dml` and `spatial_batch_norm.dml` to `batch_norm1d.dml` and `batch_norm2d.dml`
Extract preprocessing code into a Python package + script
Add `epochs` parameter to `mnist_lenet::train(...)` function.
Wrong codegen materialization decisions due to dangling hop parents
Missing common subexpression elimination on codegen cplan construction
Inconsistent generated code compared to existing builtin functions
Improved literal replacement nrow/ncol
Update Conv & Max Pooling layer names to include "2D"
General code cleanup
Potential bug in compare method
Support rowMins/rowMaxs in row aggregate template
Support 'replace' in row aggregate and cell templates
Multi-threaded compilation of fused operators
Add support for matrix-vector GPU axpy operation
Reduce the JVM memory required for transfering Numpy array
Vector primitives for row comparisons
Fuse row aggregate w/ colvector output into cell template
Support min/max/sumsq in cell templates w/ aggregation