/*=========================================================================

Program:   Visualization Toolkit
Module:    vtkMultiCorrelativeStatistics.h

Copyright (c) Ken Martin, Will Schroeder, Bill Lorensen
All rights reserved.
See Copyright.txt or http://www.kitware.com/Copyright.htm for details.

This software is distributed WITHOUT ANY WARRANTY; without even
the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
PURPOSE.  See the above copyright notice for more information.

=========================================================================*/
/*-------------------------------------------------------------------------
  Copyright 2010 Sandia Corporation.
  Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
  the U.S. Government retains certain rights in this software.
  -------------------------------------------------------------------------*/
/**
 * @class   vtkMultiCorrelativeStatistics
 * @brief   A class for multivariate linear correlation
 *
 *
 * Given a selection of sets of columns of interest, this class provides the
 * following functionalities, depending on the operation in which it is executed:
 * * Learn: calculates means, unbiased variance and covariance estimators of
 *   column pairs coefficient.
 *   More precisely, Learn calculates the averages and centered
 *   variance/covariance sums; if \p finalize is set to true (default),
 *   the final statistics are calculated.
 *   The output metadata on port OUTPUT_MODEL is a multiblock dataset containing at a minimum
 *   one vtkTable holding the raw sums in a sparse matrix style. If \a finalize is
 *   true, then one additional vtkTable will be present for each requested set of
 *   column correlations. These additional tables contain column averages, the
 *   upper triangular portion of the covariance matrix (in the upper right hand
 *   portion of the table) and the Cholesky decomposition of the covariance matrix
 *   (in the lower portion of the table beneath the covariance triangle).
 *   The leftmost column will be a vector of column averages.
 *   The last entry in the column averages vector is the number of samples.
 *   As an example, consider a request for a 3-column correlation with columns
 *   named ColA, ColB, and ColC.
 *   The resulting table will look like this:
 *   <pre>
 *      Column  |Mean     |ColA     |ColB     |ColC
 *      --------+---------+---------+---------+---------
 *      ColA    |avg(A)   |cov(A,A) |cov(A,B) |cov(A,C)
 *      ColB    |avg(B)   |chol(1,1)|cov(B,B) |cov(B,C)
 *      ColC    |avg(C)   |chol(2,1)|chol(2,2)|cov(C,C)
 *      Cholesky|length(A)|chol(3,1)|chol(3,2)|chol(3,3)
 *   </pre>
 *   The mean point and the covariance matrix can be replaced by the median point and the
 *   MAD matrix (Median Absolute Deviation) thanks to the MedianAbsoluteDeviation boolean.
 *   In this mode, the resulting table will look like this:
 *   <pre>
 *      Column  |Mean     |ColA     |ColB     |ColC
 *      --------+---------+---------+---------+---------
 *      ColA    |med(A)   |MAD(A,A) |MAD(A,B) |MAD(A,C)
 *      ColB    |med(B)   |chol(1,1)|MAD(B,B) |MAD(B,C)
 *      ColC    |med(C)   |chol(2,1)|chol(2,2)|MAD(C,C)
 *      Cholesky|length(A)|chol(3,1)|chol(3,2)|chol(3,3)
 *   </pre>
 *   The Median Absolute Deviation is known to be more robust than the covariance. It is
 *   used in the robust PCA computation for instance.
 * * Assess: given a set of results matrices as specified above in input port INPUT_MODEL and
 *   tabular data on input port INPUT_DATA that contains column names matching those
 *   of the tables on input port INPUT_MODEL, the assess mode computes the relative
 *   deviation of each observation in port INPUT_DATA's table according to the linear
 *   correlations implied by each table in port INPUT_MODEL.
 *
 * @par Thanks:
 * Thanks to Philippe Pebay, Jackson Mayo, and David Thompson of
 * Sandia National Laboratories for implementing this class.
 * Updated by Philippe Pebay, Kitware SAS 2012
 * Updated by Tristan Coulange and Joachim Pouderoux, Kitware SAS 2013
 */

#ifndef vtkMultiCorrelativeStatistics_h
#define vtkMultiCorrelativeStatistics_h

#include "vtkFiltersStatisticsModule.h" // For export macro
#include "vtkStatisticsAlgorithm.h"

class vtkDoubleArray;
class vtkMultiBlockDataSet;
class vtkOrderStatistics;
class vtkVariant;

class VTKFILTERSSTATISTICS_EXPORT vtkMultiCorrelativeStatistics : public vtkStatisticsAlgorithm
{
public:
  vtkTypeMacro(vtkMultiCorrelativeStatistics, vtkStatisticsAlgorithm);
  void PrintSelf(ostream& os, vtkIndent indent) override;
  static vtkMultiCorrelativeStatistics* New();

  /**
   * Given a collection of models, calculate aggregate model
   */
  void Aggregate(vtkDataObjectCollection*, vtkMultiBlockDataSet*) override;

  //@{
  /**
   * If set to true, the covariance matrix is replaced by
   * the Median Absolute Deviation matrix.
   * Default is false.
   */
  vtkSetMacro(MedianAbsoluteDeviation, bool);
  vtkGetMacro(MedianAbsoluteDeviation, bool);
  vtkBooleanMacro(MedianAbsoluteDeviation, bool);
  //@}

protected:
  vtkMultiCorrelativeStatistics();
  ~vtkMultiCorrelativeStatistics() override;

  /**
   * Execute the calculations required by the Learn option.
   */
  void Learn(vtkTable*, vtkTable*, vtkMultiBlockDataSet*) override;

  /**
   * Execute the calculations required by the Derive option.
   */
  void Derive(vtkMultiBlockDataSet*) override;

  /**
   * Execute the calculations required by the Assess option.
   */
  void Assess(vtkTable*, vtkMultiBlockDataSet*, vtkTable*) override;

  /**
   * Execute the calculations required by the Test option.
   */
  void Test(vtkTable*, vtkMultiBlockDataSet*, vtkTable*) override { return; }

  /**
   * Provide the appropriate assessment functor.
   */
  void SelectAssessFunctor(vtkTable* inData, vtkDataObject* inMeta, vtkStringArray* rowNames,
    AssessFunctor*& dfunc) override;

  /**
   * Computes the median of inData with vtkOrderStatistics.
   */
  virtual void ComputeMedian(vtkTable* inData, vtkTable* outData);

  /**
   * Return a new vtkOrderStatistics instance.
   * Used by derived class to return a derivate class instead.
   */
  virtual vtkOrderStatistics* CreateOrderStatisticsInstance();

  bool MedianAbsoluteDeviation;

private:
  vtkMultiCorrelativeStatistics(const vtkMultiCorrelativeStatistics&) = delete;
  void operator=(const vtkMultiCorrelativeStatistics&) = delete;
};

#endif