ignite-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexey Zinoviev <zaleslaw....@gmail.com>
Subject [Discussion] [ML] Future of Pipelines
Date Tue, 12 Feb 2019 11:14:40 GMT
Hi, Igniters! 

Currently, we have in Ignite ML next Pipeline API in alpha version via
PipelineMDL and Pipeline classes.

I'm going to finish Pipeline API for the next release 2.8 and need you help
in brainstorming

You could find an example in master  here
<https://github.com/apache/ignite/blob/master/examples/src/main/java/org/apache/ignite/examples/ml/tutorial/Step_5_Scaling_with_Pipeline.java>
 

The snippet is added here

 PipelineMdl<Integer, Object[]> mdl = new Pipeline<Integer, Object[],
Object[]>()
                    .addFeatureExtractor(featureExtractor)
                    .addLabelExtractor(lbExtractor)
                    .addPreprocessor(new EncoderTrainer<Integer, Object[]>()
                        .withEncoderType(EncoderType.STRING_ENCODER)
                        .withEncodedFeature(1)
                        .withEncodedFeature(6))
                    .addPreprocessor(new ImputerTrainer<Integer,
Object[]>())
                    .addPreprocessor(new MinMaxScalerTrainer<Integer,
Object[]>())
                    .addPreprocessor(new NormalizationTrainer<Integer,
Object[]>()
                        .withP(1))
                    .addTrainer(new DecisionTreeClassificationTrainer(5, 0))
                    .fit(ignite, dataCache);

If you have any experience with Pipeline APi in another ML framework as
scikit-learn, Spark, .NET ML and etc, please have a look at this API and
write any suggestion about possible features and use-cases

Feel free to add any suggestions or comments.




--
Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/

Mime
View raw message