Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

This FLIP proposes quite a few changes and addition to the existing Flink ML APIs. We first describe the final APIs of the classes updated by this FLIP, followed by list the detailed explanation of changes we have made to the Flink ML APIs.


The following code block shows the interface of Stage, Transformer, Estimator, Pipeline and PipelineModel after the proposed changes.

Code Block
languagejava
@PublicEvolving
interface Stage<T extends Stage<T>> extends WithParams<T>, Serializable {

    /**
     * This method checks the compatibility between input schemas, stage's parameters and stage's
     * logic. It should raise an exception if there is any mismatch, e.g. the number of input
     * schemas is wrong, or if a required field is missing from a schema.
     *
     * <p>If there is no mismatch, the method derives and returns the output schemas from the input
     * schemas.
     *
     * <p>Note that the output schemas of a given Estimator instance should equal the output schemas
     * of the Transformer instance fitted by this Estimator instance, suppose the same list of input
     * schemas are used as inputs to the fit/transform methods respectively.
     *
     * @param schemas the list of schemas of the input tables.
     * @return the list of schemas of the output tables.
     */
    TableSchema[] transformSchemas(TableSchema... schemas);

    /** Skipped */
    default String toJson() {...}

    /** Skipped */
    default void loadJson(String json) {...}
}


@PublicEvolving
public interface Transformer<T extends Transformer<T>> extends Stage<T> {

    /**
     * Applies the Transformer on the given input tables, and returns the result tables.
     *
     * @param inputs a list of tables
     * @return a list of tables
     */
    Table[] transform(Table... inputs);

    /**
     * Uses the given list of tables to update internal states. This can be useful for e.g. online
     * learning where an Estimator fits an infinite stream of training samples and streams the model
     * diff data to this Transformer.
     *
     * <p>This method may be called at most once.
     *
     * @param inputs a list of tables
     */
    default void setStateStreams(Table... inputs) {
        throw new UnsupportedOperationException("this method is not implemented");
    }

    /**
     * Gets a list of tables representing changes of internal states of this Transformer. These
     * tables might come from the Estimator that instantiated this Transformer.
     *
     * @return a list of tables
     */
    default Table[] getStateStreams() {
        throw new UnsupportedOperationException("this method is not implemented");
    }
}


@PublicEvolving
public interface Estimator<E extends Estimator<E, M>, M extends Transformer<M>> extends Stage<E> {

    /**
     * Trains on the given inputs and produces a Transformer. If this Estimator may be used to
     * compose a Pipeline, the transform method of the returned Transformer should be able to accept
     * a list of tables of the same length and schemas as the fit method of this Estimator.
     *
     * @param inputs a list of tables
     * @return a Transformer
     */
    M fit(Table... inputs);
}

@PublicEvolving
public final class Pipeline implements Estimator<Pipeline, PipelineModel> {

    public Pipeline(List<Stage<?>> stages) {...}

    @Override
    public PipelineModel fit(Table... inputs) {...}

    /** Skipped a few public APIs and implementations of the Estimator APIs. */
}


@PublicEvolving
public final class PipelineModel implements Transformer<PipelineModel> {

    public PipelineModel(List<Transformer<?>> transformers) {...}

    /** Skipped a few public APIs and implementations of the Transformer APIs. */
}



The following code block shows the interface of Graph, GraphModel and GraphBuilder proposed by this FLIP.


Code Block
languagejava
/**
 * A Graph acts as an Estimator. It consists of a DAG of stages, each of which is either an
 * Estimator or Transformer.
 */
@PublicEvolving
public final class Graph implements Estimator<Graph, GraphModel> {
    public Graph(...) {...}

    @Override
    public GraphModel fit(Table... inputs) {...}

    @Override
    public TableSchema[] transformSchemas(TableSchema... schemas) {
        return schemas;
    }

    /** Skipped a few public APIs and implementations of the Estimator APIs. */
}





Proposed Changes

Describe the new thing you want to do in appropriate detail. This may be fairly extensive and have large subsections of its own. Or it may be a few sentences. Use judgement based on the scope of the change.

...