mleap
mleap copied to clipboard
TrueCar
MLeap allows for easily putting Spark ML pipelines into production
Would be nice to support NaiveBayes classifier. It's one of the most useful classifiers for text learning.
We will need a custom Spark implementation of this, as the current one is metadata-dependent. 1. Implement Spark estimator/transformer 2. Core OneHotEncoder 3. OneHotEncoder transformer 4. Spark converter 5. Serializer...
Let's get Travis CI automatically running our tests so that we don't accidentally break things! run tests against the develop branch if we have to specify
Add test coverage for all untested components in mleap-runtime
Add in full support for LogisticRegression including: core support runtime support with a transformer and estimator Spark support with converters for estimator/transformer
1. Core model for k-means clustering 2. Transformer for k-means 3. Spark converter 4. MLeap serializer 5. Spark serializer
This will enable multinomial logistic regression and multinomial support vector machines. 1. Add core model, need to create a Classifier trait to indicate Classifier models 2. Make sure core model...
There are warnings given when compiling about the netlib native integration libraries, investigate why these are happening.
We would like the following components for DecisionTreeRegression: 1. DecisionTreeRegressionModel in mleap-runtime (this is the actual MLeap transformer) 2. Converter from Spark to MLeap in mleap-spark 3. Let's discuss serialization...