Public H2O 2 [OLD]
Back to project
Issues and filters
View all filters
Public H2O 2 [OLD]
PCA: gives AIOOBE while scoring on Airlines dataset
pca predict on test data throws "Cannot process more than 5000 columns"
h2o.deeplearning : grid search displays weird model key
R auto-transpose of vectors is not supported for H2O columns
h2o.ddply: aggregates have inconsistent handling of missing values
There is no way to unlock the keys from R.
Export predicted results to HDFS in part files
There is no way to set the seed for h2o.splitframe from R
Parsing S3 bucket fails, parsing individual files from bucket passes
Ensemble Wrappers in R
R throws a warning when printing H2OTable objects
don't see any drf json api at all yet in h2o-dev?
don't see any speeddrf json api at all in h2o dev yet?
no model metrics yet (unimplemented) for word2vec?
Documenting loading and saving models through R
h2o cluster names are not unique enough in gradle builds, to avoid unexpected cluster formations
Introduce Parquet support into H2O
Epic for Product Documentation
CoxPH: ("start", "stop", "status") not equivalent to ("elapsed", "status")
Write steam go/no go test
Grid syntax: All algos flash this msg when try to run a grid job: Field 'ntrees' : Value '5,2' is not a valid long integer.
Java tutorials for users
SRF javapredict runit failure
Add GPGPU support for H2O Deep Learning
Overarching epic for Developer Documentation
h2o probably should support ipv6 addresses. And ERROR not WARN.
Make GLM handle more than 10000 predictors with non-zero weights.
DRF2 on jacobi 6, when run on 2 machines, throws OOM error while runs fine on 1 machine with a much smaller heap size
SRF grid reports wrong prediction errors on the grid page
Try to explore option proposed by Spark guys
R: Missing documentation on s4 if else method
Bad ast created by R after deeplearning predict. seems to be missing the key for a row slice operation
impute with method=mode on enum col with NAs. seems to say there are no NAs
can't use exec to set an enum col? (trying to set to NA)
2/InsertMissingValues doesn't, if col is enum. This is bad if test assumes it's getting NAs but it's not
exec gets slower with multiple iterations
cd5 with h2o installed as service. Don't see a way to configure h2o java -jar params, notably heap size
assert isConstantResponse() fail
-jks_pass supplied password not working
MLlib backend based on the top of H2O
Sparkling Water : Issue splitting the data into training and testing
Add weighted regression analysis for GLM
Non-negative Matrix Factorization for Recommendations
Grid search not visible in browser for ntrees for GBM
RTE using Key.make()
Invoking APIs before dataset is fully parsed causes NPE
Ability to set loss function by column
Error message while running GBM
Null pointer exception - R h2o package sub/gsub functions
Compilation failed with Java 1.8
issue 1 of 903