MaxEnt.cpp

Go to the documentation of this file.

// Mantid Repository : https://github.com/mantidproject/mantid
//
// Copyright © 2018 ISIS Rutherford Appleton Laboratory UKRI,
//   NScD Oak Ridge National Laboratory, European Spallation Source,
//   Institut Laue - Langevin & CSNS, Institute of High Energy Physics, CAS
// SPDX - License - Identifier: GPL - 3.0 +
#include "MantidAlgorithms/MaxEnt.h"
#include "MantidAPI/EqualBinSizesValidator.h"
#include "MantidAPI/MatrixWorkspace.h"
#include "MantidAPI/TextAxis.h"
#include "MantidAlgorithms/MaxEnt/MaxentEntropyNegativeValues.h"
#include "MantidAlgorithms/MaxEnt/MaxentEntropyPositiveValues.h"
#include "MantidAlgorithms/MaxEnt/MaxentSpaceComplex.h"
#include "MantidAlgorithms/MaxEnt/MaxentSpaceReal.h"
#include "MantidAlgorithms/MaxEnt/MaxentTransformFourier.h"
#include "MantidAlgorithms/MaxEnt/MaxentTransformMultiFourier.h"
#include "MantidDataObjects/TableWorkspace.h"
#include "MantidDataObjects/Workspace2D.h"
#include "MantidDataObjects/WorkspaceCreation.h"
#include "MantidHistogramData/Histogram.h"
#include "MantidHistogramData/HistogramBuilder.h"
#include "MantidHistogramData/LinearGenerator.h"
#include "MantidKernel/BoundedValidator.h"
#include "MantidKernel/EqualBinsChecker.h"
#include "MantidKernel/ListValidator.h"
#include "MantidKernel/UnitFactory.h"
#include "MantidKernel/VectorHelper.h"
#include <algorithm>
#include <gsl/gsl_linalg.h>
#include <numeric>
 
namespace Mantid::Algorithms {
 
using Mantid::HistogramData::LinearGenerator;
using Mantid::HistogramData::Points;
 
using namespace API;
using namespace Kernel;
using namespace Mantid::DataObjects;
using namespace Mantid::HistogramData;
 
// Register the algorithm into the AlgorithmFactory
DECLARE_ALGORITHM(MaxEnt)
 
namespace {
 
// Maps defining the inverse caption and label for the reconstructed image
// Example:
// The input workspaces (X axis) is in (Time, s)
// The output image should be in (Frequency, Hz)
 
// Defines the new caption
std::map<std::string, std::string> inverseCaption = {
    {"Time", "Frequency"}, {"Frequency", "Time"}, {"d-Spacing", "q"}, {"q", "d-Spacing"}};
// Defines the new label
std::map<std::string, std::string> inverseLabel = {{"s", "Hz"},
                                                   {"microsecond", "MHz"},
                                                   {"Hz", "s"},
                                                   {"MHz", "microsecond"},
                                                   {"Angstrom", "Angstrom^-1"},
                                                   {"Angstrom^-1", "Angstrom"}};
// A threshold for small singular values
const double THRESHOLD = 1E-6;
 
const double BIN_WIDTH_ERROR_LEVEL = 0.5;
 
MatrixWorkspace_sptr removeZeros(MatrixWorkspace_sptr &ws, const std::vector<size_t> &itCount,
                                 const std::string &yLabel) {
 
  ws->setYUnitLabel(yLabel);
  ws->getAxis(0)->unit() = UnitFactory::Instance().create("Label");
  Unit_sptr unit = ws->getAxis(0)->unit();
  std::shared_ptr<Units::Label> label = std::dynamic_pointer_cast<Units::Label>(unit);
  label->setLabel("Number of Iterations", "");
 
  const size_t nspec = ws->getNumberHistograms();
  if (itCount.empty()) {
    return ws; // In case, we don't have any spectra
  }
  for (size_t spec = 0; spec < nspec; spec++) {
    // the workspace holds point data, so this resizes X, Y and E alike
    ws->resizeHistogram(spec, itCount[spec]);
  }
  return ws;
}
} // namespace
 
//----------------------------------------------------------------------------------------------
 
const std::string MaxEnt::name() const { return "MaxEnt"; }
 
int MaxEnt::version() const { return 1; }
 
const std::string MaxEnt::category() const { return "Arithmetic\\FFT"; }
 
const std::string MaxEnt::summary() const {
  return "Runs Maximum Entropy method on every spectrum of an input workspace. "
         "It currently works for the case where data and image are related by a"
         " 1D Fourier transform.";
}
 
//----------------------------------------------------------------------------------------------
void MaxEnt::init() {
 
  // X values in input workspace must be (almost) equally spaced
  declareProperty(
      std::make_unique<WorkspaceProperty<>>("InputWorkspace", "", Direction::Input,
                                            std::make_shared<EqualBinSizesValidator>(BIN_WIDTH_ERROR_LEVEL)),
      "An input workspace.");
 
  declareProperty("ComplexData", false,
                  "If true, the input data is assumed to be complex and the "
                  "input workspace is expected to have an even number of "
                  "histograms (2N). Spectrum numbers S and S+N are assumed to "
                  "be the real and imaginary part of the complex signal "
                  "respectively.");
 
  declareProperty("ComplexImage", true,
                  "If true, the algorithm will use complex images for the "
                  "calculations. This is the recommended option when there is "
                  "no prior knowledge about the image. If the image is known "
                  "to be real, this option can be set to false and the "
                  "algorithm will only consider the real part for "
                  "calculations.");
 
  declareProperty("PositiveImage", false,
                  "If true, the reconstructed image is only allowed to take "
                  "positive values. It can take negative values otherwise. "
                  "This option defines the entropy formula that will be used "
                  "for the calculations (see next section for more details).");
 
  declareProperty("AutoShift", false,
                  "Automatically calculate and apply phase shift. Zero on the "
                  "X axis is assumed to be in the first bin. If it is not, "
                  "setting this property will automatically correct for this.");
 
  auto mustBePositive = std::make_shared<BoundedValidator<size_t>>();
  mustBePositive->setLower(0);
  declareProperty(std::make_unique<PropertyWithValue<size_t>>("ResolutionFactor", 1, mustBePositive, Direction::Input),
                  "An integer number indicating the factor by which the number "
                  "of points will be increased in the image and reconstructed "
                  "data");
 
  auto mustBeNonNegative = std::make_shared<BoundedValidator<double>>();
  mustBeNonNegative->setLower(1E-12);
  declareProperty(std::make_unique<PropertyWithValue<double>>("A", 0.4, mustBeNonNegative, Direction::Input),
                  "A maximum entropy constant. This algorithm was first developed for the "
                  "ISIS muon group where the default 0.4 was found to give good "
                  "reconstructions. "
                  "In general the user will need to experiment with this value. Choosing a "
                  "small value may lead to unphysical spiky reconstructions and choosing "
                  "an increasingly large "
                  "value the reconstruction will start to resamble that of a direct "
                  "fourier "
                  "transform reconstruction. However, where the data contain a "
                  "zero Fourier data point with a small error the "
                  "reconstruction will be insensitive to the choice "
                  "of this property (and increasing so the more well determined "
                  "this data point is).");
 
  declareProperty(
      std::make_unique<PropertyWithValue<double>>("ChiTargetOverN", 1.0, mustBeNonNegative, Direction::Input),
      "Target value of Chi-square divided by the number of data points (N)");
 
  declareProperty(std::make_unique<PropertyWithValue<double>>("ChiEps", 0.001, mustBeNonNegative, Direction::Input),
                  "Required precision for Chi-square");
 
  declareProperty(
      std::make_unique<PropertyWithValue<double>>("DistancePenalty", 0.1, mustBeNonNegative, Direction::Input),
      "Distance penalty applied to the current image at each iteration.");
 
  declareProperty(std::make_unique<PropertyWithValue<double>>("MaxAngle", 0.001, mustBeNonNegative, Direction::Input),
                  "Maximum degree of non-parallelism between S (the entropy) and C "
                  "(chi-squared). These needs to be parallel. Chosing a smaller "
                  "shouldn't change the output. However, if you find this is the "
                  "case please let the Mantid team know since this indicates that "
                  "the default value of this proporty may need changing or "
                  "other changes to this implementation are required.");
 
  mustBePositive = std::make_shared<BoundedValidator<size_t>>();
  mustBePositive->setLower(1);
  declareProperty(std::make_unique<PropertyWithValue<size_t>>("MaxIterations", 20000, mustBePositive, Direction::Input),
                  "Maximum number of iterations.");
 
  declareProperty(
      std::make_unique<PropertyWithValue<size_t>>("AlphaChopIterations", 500, mustBePositive, Direction::Input),
      "Maximum number of iterations in alpha chop.");
  declareProperty(
      std::make_unique<WorkspaceProperty<>>("DataLinearAdj", "", Direction::Input, PropertyMode::Optional,
                                            std::make_shared<EqualBinSizesValidator>(BIN_WIDTH_ERROR_LEVEL)),
      "Adjusts the calculated data by multiplying each value by the "
      "corresponding Y value of this workspace. "
      "The data in this workspace is complex in the same manner as complex "
      "input data.");
  declareProperty(
      std::make_unique<WorkspaceProperty<>>("DataConstAdj", "", Direction::Input, PropertyMode::Optional,
                                            std::make_shared<EqualBinSizesValidator>(BIN_WIDTH_ERROR_LEVEL)),
      "Adjusts the calculated data by adding to each value the corresponding Y "
      "value of this workspace. "
      "If DataLinearAdj is also specified, this addition is done after its "
      "multiplication. "
      "See equation in documentation for how DataLinearAdj and DataConstAdj "
      "are applied. "
      "The data in this workspace is complex in the same manner as complex "
      "input data.");
  declareProperty("PerSpectrumReconstruction", true,
                  "Reconstruction is done independently on each spectrum. "
                  "If false, all the spectra use one image and the reconstructions "
                  "differ only through their adjustments. "
                  "ComplexData must be set true, when this is false.");
 
  declareProperty(std::make_unique<WorkspaceProperty<>>("EvolChi", "", Direction::Output),
                  "Output workspace containing the evolution of Chi-sq.");
  declareProperty(std::make_unique<WorkspaceProperty<>>("EvolAngle", "", Direction::Output),
                  "Output workspace containing the evolution of "
                  "non-paralellism between S and C.");
  declareProperty(std::make_unique<WorkspaceProperty<>>("ReconstructedImage", "", Direction::Output),
                  "The output workspace containing the reconstructed image.");
  declareProperty(std::make_unique<WorkspaceProperty<>>("ReconstructedData", "", Direction::Output),
                  "The output workspace containing the reconstructed data.");
}
 
//----------------------------------------------------------------------------------------------
void MaxEnt::validateBinEdges(const std::string &wsName, std::map<std::string, std::string> &messages) {
  const double warningLevel = 0.01;
  MatrixWorkspace_const_sptr ws = getProperty(wsName);
  if (ws) {
 
    Kernel::EqualBinsChecker binChecker(ws->x(0), BIN_WIDTH_ERROR_LEVEL, warningLevel);
    const std::string binError = binChecker.validate();
    if (!binError.empty()) {
      messages[wsName] = binError;
    }
  }
}
 
//----------------------------------------------------------------------------------------------
std::map<std::string, std::string> MaxEnt::validateInputs() {
 
  std::map<std::string, std::string> result;
 
  validateBinEdges("InputWorkspace", result);
  validateBinEdges("DataLinearAdj", result);
  validateBinEdges("DataConstAdj", result);
 
  MatrixWorkspace_sptr inWS = getProperty("InputWorkspace");
 
  size_t nHistograms = 0;
  if (inWS) {
    // If the input signal is complex, we expect an even number of histograms
    // in the input workspace
 
    nHistograms = inWS->getNumberHistograms();
    bool complex = getProperty("ComplexData");
    if (complex && (nHistograms % 2))
      result["InputWorkspace"] = "The number of histograms in the input "
                                 "workspace must be even for complex data";
    if (!complex)
      nHistograms *= 2; // Double number of real histograms to compare with
                        // adjustments, which are always complex.
  }
 
  // Check linear adjustments, we expect and even number of histograms
  // and if any, they must be sufficient for all spectra in input workspace,
  // if per spectrum reconstruction is done.
  MatrixWorkspace_sptr linAdj = getProperty("DataLinearAdj");
  size_t nAHistograms = 0;
  if (linAdj)
    nAHistograms = linAdj->getNumberHistograms();
  if (nAHistograms % 2)
    result["DataLinearAdj"] = "The number of histograms in the linear "
                              "adjustments workspace must be even, because they are complex data";
  else if (nAHistograms > 0 && nAHistograms < nHistograms)
    result["DataLinearAdj"] = "The number of histograms in the linear "
                              "adjustments workspace is insufficient for the input workspace";
 
  // Check constant adjustments, we expect and even number of histograms
  // and if any, they must be sufficient for all spectra in input workspace,
  // if per spectrum reconstruction is done.
  MatrixWorkspace_sptr constAdj = getProperty("DataConstAdj");
  nAHistograms = 0;
  if (constAdj)
    nAHistograms = constAdj->getNumberHistograms();
  if (nAHistograms % 2)
    result["DataConstAdj"] = "The number of histograms in the constant "
                             "adjustments workspace must be even, because they are complex data";
  else if (nAHistograms > 0 && nAHistograms < nHistograms)
    result["DataConstAdj"] = "The number of histograms in the constant "
                             "adjustments workspace is insufficient for the input workspace";
 
  return result;
}
 
//----------------------------------------------------------------------------------------------
void MaxEnt::exec() {
 
  // MaxEnt parameters
  // Complex data?
  const bool complexData = getProperty("ComplexData");
  // Complex image?
  const bool complexImage = getProperty("ComplexImage");
  // Image must be positive?
  const bool positiveImage = getProperty("PositiveImage");
  // Autoshift
  const bool autoShift = getProperty("AutoShift");
  // Increase the number of points in the image by this factor
  const size_t resolutionFactor = getProperty("ResolutionFactor");
  // Background (default level, sky background, etc)
  const double background = getProperty("A");
  // Chi target
  const double ChiTargetOverN = getProperty("ChiTargetOverN");
  // Required precision for Chi arget
  const double chiEps = getProperty("ChiEps");
  // Maximum degree of non-parallelism between S and C
  const double angle = getProperty("MaxAngle");
  // Distance penalty for current image
  const double distEps = getProperty("DistancePenalty");
  // Maximum number of iterations
  const size_t nIter = getProperty("MaxIterations");
  // Maximum number of iterations in alpha chop
  const size_t alphaIter = getProperty("AlphaChopIterations");
  // Number of spectra and datapoints
  // Read input workspace
  MatrixWorkspace_const_sptr inWS = getProperty("InputWorkspace");
  // Number of spectra
  size_t nHist = inWS->getNumberHistograms();
  // Number of data points - assumed to be constant between spectra or
  // this will throw an exception
  size_t npoints = inWS->blocksize() * resolutionFactor;
  // Number of X bins
  const size_t npointsX = inWS->isHistogramData() ? npoints + 1 : npoints;
  // Linear adjustment of calculated data
  MatrixWorkspace_const_sptr dataLinearAdj = getProperty("DataLinearAdj");
  // Constant adjustment of calculated data
  MatrixWorkspace_const_sptr dataConstAdj = getProperty("DataConstAdj");
  // Add spectra in reconstruction if false
  const bool perSpectrumReconstruction = getProperty("PerSpectrumReconstruction");
 
  // For now have the requirement that data must have non-zero
  // (and positive!) errors
  for (size_t s = 0; s < nHist; s++) {
    const auto &errors = inWS->e(s);
    if (std::any_of(errors.cbegin(), errors.cend(), [](const auto error) { return error <= 0.; })) {
      throw std::invalid_argument("Input data must have all errors non-zero.");
    }
  }
 
  // Is our data space real or complex?
  MaxentSpace_sptr dataSpace;
  if (complexData) {
    dataSpace = std::make_shared<MaxentSpaceComplex>();
  } else {
    dataSpace = std::make_shared<MaxentSpaceReal>();
  }
  // Is our image space real or complex?
  MaxentSpace_sptr imageSpace;
  if (complexImage) {
    imageSpace = std::make_shared<MaxentSpaceComplex>();
  } else {
    imageSpace = std::make_shared<MaxentSpaceReal>();
  }
  // The type of transform. Currently a 1D Fourier Transform or Multiple ID
  // Fourier transform
  MaxentTransform_sptr transform;
  if (perSpectrumReconstruction) {
    transform = std::make_shared<MaxentTransformFourier>(dataSpace, imageSpace);
  } else {
    auto complexDataSpace = std::make_shared<MaxentSpaceComplex>();
    transform = std::make_shared<MaxentTransformMultiFourier>(complexDataSpace, imageSpace, nHist / 2);
  }
 
  // The type of entropy we are going to use (depends on the type of image,
  // positive only, or positive and/or negative)
  MaxentEntropy_sptr entropy;
  if (positiveImage) {
    entropy = std::make_shared<MaxentEntropyPositiveValues>();
  } else {
    entropy = std::make_shared<MaxentEntropyNegativeValues>();
  }
 
  // Entropy and transform is all we need to set up a calculator
  MaxentCalculator maxentCalculator = MaxentCalculator(entropy, transform);
 
  // Output workspaces
  MatrixWorkspace_sptr outImageWS;
  MatrixWorkspace_sptr outDataWS;
  MatrixWorkspace_sptr outEvolChi;
  MatrixWorkspace_sptr outEvolTest;
 
  size_t nDataSpec = complexData ? nHist / 2 : nHist;
  size_t nImageSpec = nDataSpec;
  size_t nSpecConcat = 1;
  if (!perSpectrumReconstruction) {
    nSpecConcat = nImageSpec;
    nImageSpec = 1;
  }
  outImageWS = create<MatrixWorkspace>(*inWS, 2 * nImageSpec, Points(npoints));
  for (size_t i = 0; i < outImageWS->getNumberHistograms(); ++i)
    outImageWS->getSpectrum(i).setDetectorID(static_cast<detid_t>(i + 1));
  HistogramBuilder builder;
  builder.setX(npointsX);
  builder.setY(npoints);
  builder.setDistribution(inWS->isDistribution());
  outDataWS = create<MatrixWorkspace>(*inWS, 2 * nDataSpec, builder.build());
 
  for (size_t i = 0; i < outDataWS->getNumberHistograms(); ++i)
    outDataWS->getSpectrum(i).setDetectorID(static_cast<detid_t>(i + 1));
  outEvolChi = create<MatrixWorkspace>(*inWS, nImageSpec, Points(nIter));
  outEvolTest = create<MatrixWorkspace>(*inWS, nImageSpec, Points(nIter));
 
  npoints = complexImage ? npoints * 2 : npoints;
  std::vector<size_t> iterationCounts;
  iterationCounts.reserve(nImageSpec);
  outEvolChi->setPoints(0, Points(nIter, LinearGenerator(0.0, 1.0)));
 
  size_t dataLength = complexData ? 2 * inWS->y(0).size() : inWS->y(0).size();
  dataLength *= nSpecConcat;
 
  for (size_t spec = 0; spec < nImageSpec; spec++) {
 
    // Start distribution (flat background)
    std::vector<double> image(npoints, background);
 
    std::vector<double> data(dataLength, 0.0);
    std::vector<double> errors(dataLength, 0.0);
    if (complexData) {
      data = toComplex(inWS, spec, false,
                       !perSpectrumReconstruction); // 3rd arg false -> data
      errors = toComplex(inWS, spec, true,
                         !perSpectrumReconstruction); // 3rd arg true -> errors
    } else {
      if (!perSpectrumReconstruction) {
        throw std::invalid_argument("ComplexData must be true, if PerSpectrumReconstruction is false.");
      } else {
        data = inWS->y(spec).rawData();
        errors = inWS->e(spec).rawData();
      }
    }
 
    std::vector<double> linearAdjustments;
    std::vector<double> constAdjustments;
    if (dataLinearAdj) {
      linearAdjustments = toComplex(dataLinearAdj, spec, false, !perSpectrumReconstruction);
    }
    if (dataConstAdj) {
      constAdjustments = toComplex(dataConstAdj, spec, false, !perSpectrumReconstruction);
    }
 
    // To record the algorithm's progress
    std::vector<double> evolChi(nIter, 0.);
    std::vector<double> evolTest(nIter, 0.);
 
    // Progress
    Progress progress(this, 0.0, 1.0, nIter);
 
    // Run maxent algorithm
    bool converged = false;
    for (size_t it = 0; it < nIter; it++) {
 
      // Iterates one step towards the solution. This means calculating
      // quadratic coefficients, search directions, angle and chi-sq
      maxentCalculator.iterate(data, errors, image, background, linearAdjustments, constAdjustments);
 
      // Calculate delta to construct new image (SB eq. 25)
      double currChisq = maxentCalculator.getChisq();
      auto coeffs = maxentCalculator.getQuadraticCoefficients();
      auto delta = move(coeffs, ChiTargetOverN / currChisq, chiEps, alphaIter);
 
      // Apply distance penalty (SB eq. 33)
      delta = applyDistancePenalty(delta, coeffs, image, background, distEps);
 
      // Update image
      auto dirs = maxentCalculator.getSearchDirections();
      image = updateImage(image, delta, dirs);
 
      // Record the evolution of Chi-square and angle(S,C)
      double currAngle = maxentCalculator.getAngle();
      evolChi[it] = currChisq;
      evolTest[it] = currAngle;
 
      // Stop condition for convergence, solution found
      if ((std::abs(currChisq / ChiTargetOverN - 1.) < chiEps) && (currAngle < angle)) {
 
        // it + 1 iterations have been done because we count from zero
        g_log.information() << "Converged after " << it + 1 << " iterations" << std::endl;
        iterationCounts.emplace_back(it + 1);
        converged = true;
        break;
      }
 
      // Check for canceling the algorithm
      if (!(it % 1000)) {
        interruption_point();
      }
 
      progress.report();
 
    } // Next Iteration
 
    // If we didn't converge, we still need to record the number of iterations
    if (!converged) {
      iterationCounts.emplace_back(nIter);
    }
 
    // Get calculated data
    auto solData = maxentCalculator.getReconstructedData();
    auto solImage = maxentCalculator.getImage();
 
    // Populate the output workspaces
    populateDataWS(inWS, spec, nDataSpec, solData, !perSpectrumReconstruction, complexData, outDataWS);
    populateImageWS(inWS, spec, nImageSpec, solImage, complexImage, outImageWS, autoShift);
 
    // Populate workspaces recording the evolution of Chi and Test
    // X values
    outEvolChi->setSharedX(spec, outEvolChi->sharedX(0));
    outEvolTest->setSharedX(spec, outEvolChi->sharedX(0));
 
    // Y values
    outEvolChi->setCounts(spec, std::move(evolChi));
    outEvolTest->setCounts(spec, std::move(evolTest));
    // No errors
 
  } // Next spectrum
  setProperty("EvolChi", removeZeros(outEvolChi, iterationCounts, "Chi squared"));
  setProperty("EvolAngle", removeZeros(outEvolTest, iterationCounts, "Maximum Angle"));
  setProperty("ReconstructedImage", outImageWS);
  setProperty("ReconstructedData", outDataWS);
}
 
//----------------------------------------------------------------------------------------------
 
std::vector<double> MaxEnt::toComplex(API::MatrixWorkspace_const_sptr &inWS, size_t spec, bool errors,
                                      bool concatSpec) {
  const size_t numBins = inWS->y(0).size();
  size_t nSpec = inWS->getNumberHistograms() / 2;
  std::vector<double> result;
  result.reserve(2 * numBins);
 
  if (inWS->getNumberHistograms() % 2)
    throw std::invalid_argument("Cannot convert input workspace to complex data");
 
  size_t nSpecOfInterest = (concatSpec ? nSpec : 1);
  size_t firstSpecOfInterest = (concatSpec ? 0 : spec);
 
  for (size_t s = firstSpecOfInterest; s < firstSpecOfInterest + nSpecOfInterest; s++) {
    if (!errors) {
      for (size_t i = 0; i < numBins; i++) {
        result.emplace_back(inWS->y(s)[i]);
        result.emplace_back(inWS->y(s + nSpec)[i]);
      }
    } else {
      for (size_t i = 0; i < numBins; i++) {
        result.emplace_back(inWS->e(s)[i]);
        result.emplace_back(inWS->e(s + nSpec)[i]);
      }
    }
  }
 
  return result;
}
 
std::vector<double> MaxEnt::move(const QuadraticCoefficients &coeffs, double ChiTargetOverN, double chiEps,
                                 size_t alphaIter) {
 
  double aMin = 0.; // Minimum alpha
  double aMax = 1.; // Maximum alpha
 
  // Dimension, number of search directions
  size_t dim = coeffs.c2.size().first;
 
  std::vector<double> deltaMin(dim, 0); // delta at alpha min
  std::vector<double> deltaMax(dim, 0); // delta at alpha max
 
  double chiMin = calculateChi(coeffs, aMin, deltaMin); // Chi at alpha min
  double chiMax = calculateChi(coeffs, aMax, deltaMax); // Chi at alpha max
 
  double dchiMin = chiMin - ChiTargetOverN; // max - target
  double dchiMax = chiMax - ChiTargetOverN; // min - target
 
  if (dchiMin * dchiMax > 0) {
    // ChiTargetOverN could be outside the range [chiMin, chiMax]
 
    if (fabs(dchiMin) < fabs(dchiMax)) {
      return deltaMin;
    } else {
      return deltaMax;
    }
    //    throw std::runtime_error("Error in alpha chop\n");
  }
 
  // Initial values of eps and iter to start while loop
  double eps = 2. * chiEps;
  size_t iter = 0;
 
  // Bisection method
 
  std::vector<double> delta(dim, 0); // delta at current alpha
 
  while ((fabs(eps / ChiTargetOverN) > chiEps) && (iter < alphaIter)) {
 
    double aMid = 0.5 * (aMin + aMax);
    double chiMid = calculateChi(coeffs, aMid, delta);
 
    eps = chiMid - ChiTargetOverN;
 
    if (dchiMin * eps > 0) {
      aMin = aMid;
      dchiMin = eps;
    }
 
    if (dchiMax * eps > 0) {
      aMax = aMid;
      dchiMax = eps;
    }
 
    iter++;
  }
 
  // Check if move was successful
  if ((fabs(eps / ChiTargetOverN) > chiEps) || (iter > alphaIter)) {
    throw std::runtime_error("Error encountered when calculating solution "
                             "image. No convergence in alpha chop.\n");
  }
 
  return delta;
}
 
double MaxEnt::calculateChi(const QuadraticCoefficients &coeffs, double a, std::vector<double> &b) {
 
  size_t dim = coeffs.c2.size().first;
 
  double ax = a;
  double bx = 1 - ax;
 
  Kernel::DblMatrix A(dim, dim);
  Kernel::DblMatrix B(dim, 1);
 
  // Construct the matrix A and vector B such that Ax=B
  for (size_t k = 0; k < dim; k++) {
    for (size_t l = 0; l < dim; l++) {
      A[k][l] = bx * coeffs.c2[k][l] - ax * coeffs.s2[k][l];
    }
    B[k][0] = -bx * coeffs.c1[k][0] + ax * coeffs.s1[k][0];
  }
 
  // Alternatives I have tried:
  // Gauss-Jordan
  // LU
  // SVD seems to work better
 
  // Solve using SVD
  b = solveSVD(A, B);
 
  // Now compute Chi
  double ww = 0.;
  for (size_t k = 0; k < dim; k++) {
    double z = 0.;
    for (size_t l = 0; l < dim; l++) {
      z += coeffs.c2[k][l] * b[l];
    }
    ww += b[k] * (coeffs.c1[k][0] + 0.5 * z);
  }
 
  // Return chi
  return ww + 1.;
}
 
std::vector<double> MaxEnt::solveSVD(DblMatrix &A, const DblMatrix &B) {
 
  size_t dim = A.size().first;
 
  auto a = gsl_matrix_view_array(A[0], dim, dim);
  auto b = gsl_vector_const_view_array(B[0], dim);
 
  std::vector<double> vVec(dim * dim), sVec(dim), wVec(dim), delta(dim);
 
  auto v = gsl_matrix_view_array(vVec.data(), dim, dim);
  auto s = gsl_vector_view_array(sVec.data(), dim);
  auto w = gsl_vector_view_array(wVec.data(), dim);
  auto x = gsl_vector_view_array(delta.data(), dim);
 
  // Singular value decomposition
  gsl_linalg_SV_decomp(&a.matrix, &v.matrix, &s.vector, &w.vector);
 
  // A could be singular or ill-conditioned. We can use SVD to obtain a least
  // squares solution by setting the small (compared to the maximum) singular
  // values to zero
 
  // Find largest sing value
  double max = *std::max_element(sVec.begin(), sVec.end());
 
  // Apply a threshold to small singular values
  double threshold = THRESHOLD * max;
  std::transform(sVec.begin(), sVec.end(), sVec.begin(), [&threshold](double el) { return el > threshold ? el : 0.0; });
 
  // Solve A*x = B
  gsl_linalg_SV_solve(&a.matrix, &v.matrix, &s.vector, &b.vector, &x.vector);
 
  return delta;
}
 
std::vector<double> MaxEnt::applyDistancePenalty(const std::vector<double> &delta, const QuadraticCoefficients &coeffs,
                                                 const std::vector<double> &image, double background, double distEps) {
 
  const double pointSum = std::accumulate(image.cbegin(), image.cend(), 0.,
                                          [](const auto sum, const auto point) { return sum + std::abs(point); });
 
  const size_t dim = coeffs.s2.size().first;
 
  double dist = 0.;
 
  for (size_t k = 0; k < dim; k++) {
    double sum = 0.0;
    for (size_t l = 0; l < dim; l++)
      sum -= coeffs.s2[k][l] * delta[l];
    dist += delta[k] * sum;
  }
 
  if (dist > distEps * pointSum / background) {
    auto newDelta = delta;
    for (size_t k = 0; k < delta.size(); k++) {
      newDelta[k] *= sqrt(distEps * pointSum / dist / background);
    }
    return newDelta;
  }
  return delta;
}
 
std::vector<double> MaxEnt::updateImage(const std::vector<double> &image, const std::vector<double> &delta,
                                        const std::vector<std::vector<double>> &dirs) {
 
  if (image.empty() || dirs.empty() || (delta.size() != dirs.size())) {
    throw std::runtime_error("Cannot calculate new image");
  }
 
  std::vector<double> newImage = image;
 
  // Calculate the new image
  for (size_t i = 0; i < image.size(); i++) {
    for (size_t k = 0; k < delta.size(); k++) {
      newImage[i] += delta[k] * dirs[k][i];
    }
  }
  return newImage;
}
 
void MaxEnt::populateImageWS(MatrixWorkspace_const_sptr &inWS, size_t spec, size_t nspec,
                             const std::vector<double> &result, bool complex, MatrixWorkspace_sptr &outWS,
                             bool autoShift) {
 
  if (complex && result.size() % 2)
    throw std::invalid_argument("Cannot write image results to output workspaces");
 
  int npoints = complex ? static_cast<int>(result.size() / 2) : static_cast<int>(result.size());
  MantidVec X(npoints);
  MantidVec YR(npoints);
  MantidVec YI(npoints);
  MantidVec E(npoints, 0.);
 
  auto dataPoints = inWS->points(spec);
  double x0 = dataPoints[0];
  double dx = dataPoints[1] - x0;
 
  double delta = 1. / dx / npoints;
  const int isOdd = (inWS->y(0).size() % 2) ? 1 : 0;
 
  double shift = x0 * 2. * M_PI;
  if (!autoShift)
    shift = 0.;
 
  // X values
  for (int i = 0; i < npoints; i++) {
    X[i] = delta * (-npoints / 2 + i);
  }
 
  // Y values
  if (complex) {
    for (int i = 0; i < npoints; i++) {
      int j = (npoints / 2 + i + isOdd) % npoints;
      double xShift = X[i] * shift;
      double c = cos(xShift);
      double s = sin(xShift);
      YR[i] = result[2 * j] * c - result[2 * j + 1] * s;
      YI[i] = result[2 * j] * s + result[2 * j + 1] * c;
      YR[i] *= dx;
      YI[i] *= dx;
    }
  } else {
    for (int i = 0; i < npoints; i++) {
      int j = (npoints / 2 + i + isOdd) % npoints;
      double xShift = X[i] * shift;
      double c = cos(xShift);
      double s = sin(xShift);
      YR[i] = result[j] * c;
      YI[i] = result[j] * s;
      YR[i] *= dx;
      YI[i] *= dx;
    }
  }
 
  // X caption & label
  auto inputUnit = inWS->getAxis(0)->unit();
  if (inputUnit) {
    std::shared_ptr<Kernel::Units::Label> lblUnit =
        std::dynamic_pointer_cast<Kernel::Units::Label>(UnitFactory::Instance().create("Label"));
    if (lblUnit) {
 
      lblUnit->setLabel(inverseCaption[inWS->getAxis(0)->unit()->caption()],
                        inverseLabel[inWS->getAxis(0)->unit()->label().ascii()]);
      outWS->getAxis(0)->unit() = lblUnit;
    }
  }
 
  outWS->mutableX(spec) = X;
  outWS->mutableY(spec) = YR;
  outWS->mutableE(spec) = E;
  outWS->setSharedX(nspec + spec, outWS->sharedX(spec));
  outWS->mutableY(nspec + spec) = YI;
  outWS->setSharedE(nspec + spec, outWS->sharedE(spec));
}
 
void MaxEnt::populateDataWS(MatrixWorkspace_const_sptr &inWS, size_t spec, size_t nspec,
                            const std::vector<double> &result, bool concatenated, bool complex,
                            MatrixWorkspace_sptr &outWS) {
 
  if (complex && result.size() % 2)
    throw std::invalid_argument("Cannot write data results to output workspaces");
  if (concatenated && !complex)
    throw std::invalid_argument("Concatenated data results must be complex");
  if (concatenated && result.size() % (nspec * 2))
    throw std::invalid_argument("Cannot write complex concatenated data results to output workspaces");
  if (concatenated && spec != 0)
    throw std::invalid_argument("Cannot write concatenated data results to "
                                "output workspaces from non-first spectrum");
 
  int resultLength = complex ? static_cast<int>(result.size() / 2) : static_cast<int>(result.size());
  size_t spectrumLength = (concatenated ? resultLength / nspec : resultLength);
  size_t spectrumLengthX = inWS->isHistogramData() ? spectrumLength + 1 : spectrumLength;
  size_t nSpecAnalyzed = (concatenated ? nspec : 1);
 
  // Here we assume equal constant binning for all spectra analyzed
  double x0 = inWS->x(spec)[0];
  double dx = inWS->x(spec)[1] - x0;
 
  // Loop over each spectrum being analyzed - one spectrum unless concatenated
  for (size_t specA = spec; specA < spec + nSpecAnalyzed; specA++) {
 
    MantidVec X(spectrumLengthX);
    MantidVec YR(spectrumLength);
    MantidVec YI(spectrumLength);
    MantidVec E(spectrumLength, 0.);
 
    // X values
    for (size_t i = 0; i < spectrumLengthX; i++) {
      X[i] = x0 + static_cast<double>(i) * dx;
    }
 
    // Y values
    if (complex) {
      if (concatenated) {
        // note the spec=0, so specA starts from 0 in this case.
        for (size_t i = 0; i < spectrumLength; i++) {
          YR[i] = result[2 * i + 2 * specA * spectrumLength];
          YI[i] = result[2 * i + 1 + 2 * specA * spectrumLength];
        }
      } else {
        for (size_t i = 0; i < spectrumLength; i++) {
          YR[i] = result[2 * i];
          YI[i] = result[2 * i + 1];
        }
      }
    } else {
      for (size_t i = 0; i < spectrumLength; i++) {
        YR[i] = result[i];
        YI[i] = 0.;
      }
    }
 
    outWS->mutableX(specA) = X;
    outWS->mutableY(specA) = YR;
    outWS->mutableE(specA) = E;
    outWS->mutableY(nspec + specA) = YI;
    outWS->setSharedX(nspec + specA, outWS->sharedX(spec));
    outWS->setSharedE(nspec + specA, outWS->sharedE(spec));
  } // Next spectrum if concatenated
}
 
} // namespace Mantid::Algorithms