[jira] [Commented] (APEXMALHAR-2094) Quantiles sketch operator

["ASF GitHub Bot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/APEXMALHAR-2094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15311082#comment-15311082 ] 

ASF GitHub Bot commented on APEXMALHAR-2094:
--------------------------------------------

Github user sandeep-n commented on a diff in the pull request:

    https://github.com/apache/incubator-apex-malhar/pull/301#discussion_r65440216
  
    --- Diff: sketches/src/main/java/org/apache/apex/malhar/sketches/QuantilesEstimator.java ---
    @@ -0,0 +1,182 @@
    +/**
    + * Licensed to the Apache Software Foundation (ASF) under one
    + * or more contributor license agreements.  See the NOTICE file
    + * distributed with this work for additional information
    + * regarding copyright ownership.  The ASF licenses this file
    + * to you under the Apache License, Version 2.0 (the
    + * "License"); you may not use this file except in compliance
    + * with the License.  You may obtain a copy of the License at
    + *
    + *   http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing,
    + * software distributed under the License is distributed on an
    + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
    + * KIND, either express or implied.  See the License for the
    + * specific language governing permissions and limitations
    + * under the License.
    + */
    +package org.apache.apex.malhar.sketches;
    +
    +import com.yahoo.sketches.quantiles.QuantilesSketch;
    +
    +import com.datatorrent.api.DefaultInputPort;
    +import com.datatorrent.api.DefaultOutputPort;
    +import com.datatorrent.api.annotation.OperatorAnnotation;
    +import com.datatorrent.common.util.BaseOperator;
    +
    +/**
    + * An implementation of BaseOperator that computes a "sketch" (a representation of the probability distribution using
    + * a low memory footprint) of the incoming numeric data, and evaluates/outputs the cumulative distribution function and
    + * quantiles of the probability distribution. Leverages the quantiles sketch implementation from the Yahoo Datasketches
    + * Library.
    + * <p/>
    + * <b>Input Port(s) : </b><br>
    + * <b>data : </b> Data values input port. <br>
    + * <br>
    + * <b>Output Port(s) : </b> <br>
    + * <b>cdfOutput : </b>cumulative distribution function output port. <br>
    + * <b>quantilesOutput : </b>quantiles output port. <br>
    + * <br>
    + * <b>Partitions : No</b>, no will yield wrong results. <br>
    + * <br>+
    + */
    +@OperatorAnnotation(partitionable = false)
    +public class QuantilesEstimator extends BaseOperator
    +{
    +  /**
    +   * Output port that emits cdf estimated at the current data point
    +   */
    +  public final transient DefaultOutputPort<Double> cdfOutput = new DefaultOutputPort<>();
    +  /**
    +   * Emits quantiles of stream seen thus far
    +   */
    +  public final transient DefaultOutputPort<double[]> quantilesOutput = new DefaultOutputPort<>();
    +  /**
    +   * Emits probability masses on specified intervals
    +   */
    +  public final transient DefaultOutputPort<double[]> pmfOutput = new DefaultOutputPort<>();
    +  private transient QuantilesSketch quantilesSketch = QuantilesSketch.builder().build();
    +  /**
    +   * This field determines the specific quantiles to be calculated.
    +   * Default is set to compute the standard quartiles.
    +   */
    +  private double[] fractions = {0.0, 0.25, 0.50, 0.75, 1.00};
    +  /**
    +   * This field determines the intervals on which the probability mass function is computed.
    +   */
    +  private double[] pmfIntervals = {};
    +  /**
    +   * This operator computes three different quantities which are output on separate output ports. If not using any of
    +   * these quantities, these variables can be set to avoid unnecessary computation.
    +   */
    +  private boolean computeCdf = true;
    +  private boolean computeQuantiles = true;
    +  private boolean computePmf = true;
    +  public final transient DefaultInputPort<Double> data = new DefaultInputPort<Double>()
    +  {
    +    @Override
    +    public void process(Double input)
    +    {
    +
    +      quantilesSketch.update(input);
    +
    +      if (computeQuantiles) {
    +        /**
    +         * Computes and emits quantiles of the stream seen thus far
    +         */
    +        quantilesOutput.emit(quantilesSketch.getQuantiles(fractions));
    +      }
    +
    +      if (computeCdf) {
    +        /**
    +         * Emits (estimate of the) cumulative distribution function evaluated at the input value, according to the
    +         * sketched probability distribution of the stream seen thus far.
    +         */
    +        cdfOutput.emit(quantilesSketch.getCDF(new double[]{input})[0]);
    +      }
    +
    +      if (computePmf) {
    +        pmfOutput.emit(quantilesSketch.getPMF(pmfIntervals));
    +      }
    +
    +    }
    +  };
    +
    +  /**
    +   * Constructor that allows non-default initialization of the quantile sketch object
    +   *
    +   * @param k:    Parameter that determines accuracy and memory usage of quantile sketch. See QuantilesSketch documentation
    --- End diff --
    
    Agreed. Unfortunately, this is the name that the underlying datasketches implementation uses, with no more descriptive name for this parameter being used either in the datasketches documentation or the research work on which this is based. "sketchSize" would be a reasonable name, but I figured it would be more straightforward for users to look up if I continued to call it k.


> Quantiles sketch operator
> -------------------------
>
>                 Key: APEXMALHAR-2094
>                 URL: https://issues.apache.org/jira/browse/APEXMALHAR-2094
>             Project: Apache Apex Malhar
>          Issue Type: New Feature
>            Reporter: Sandeep Narayanaswami
>            Assignee: Sandeep Narayanaswami
>            Priority: Minor
>
> An operator that "sketches" in an online fashion the probability distribution of an input (numeric) data stream, enabling computation of quantiles and cumulative distribution functions.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)