C++ TreeType::Point方法代码示例

void NeighborSearchRules<
    SortPolicy,
    MetricType,
    TreeType>::
UpdateAfterRecursion(TreeType& queryNode, TreeType& /* referenceNode */)
{
  // Find the worst distance that the children found (including any points), and
  // update the bound accordingly.
  double worstDistance = SortPolicy::BestDistance();

  // First look through children nodes.
  for (size_t i = 0; i < queryNode.NumChildren(); ++i)
  {
    if (SortPolicy::IsBetter(worstDistance, queryNode.Child(i).Stat().Bound()))
      worstDistance = queryNode.Child(i).Stat().Bound();
  }

  // Now look through children points.
  for (size_t i = 0; i < queryNode.NumPoints(); ++i)
  {
    if (SortPolicy::IsBetter(worstDistance,
        distances(distances.n_rows - 1, queryNode.Point(i))))
      worstDistance = distances(distances.n_rows - 1, queryNode.Point(i));
  }

  // Take the worst distance from all of these, and update our bound to reflect
  // that.
  queryNode.Stat().Bound() = worstDistance;
}

  FastMKSStat(const TreeType& node) :
      bound(-DBL_MAX),
      lastKernel(0.0),
      lastKernelNode(NULL)
  {
    // Do we have to calculate the centroid?
    if (tree::TreeTraits<TreeType>::FirstPointIsCentroid)
    {
      // If this type of tree has self-children, then maybe the evaluation is
      // already done.  These statistics are built bottom-up, so the child stat
      // should already be done.
      if ((tree::TreeTraits<TreeType>::HasSelfChildren) &&
          (node.NumChildren() > 0) &&
          (node.Point(0) == node.Child(0).Point(0)))
      {
        selfKernel = node.Child(0).Stat().SelfKernel();
      }
      else
      {
        selfKernel = sqrt(node.Metric().Kernel().Evaluate(
            node.Dataset().col(node.Point(0)),
            node.Dataset().col(node.Point(0))));
      }
    }
    else
    {
      // Calculate the centroid.
      arma::vec center;
      node.Center(center);

      selfKernel = sqrt(node.Metric().Kernel().Evaluate(center, center));
    }
  }

inline double KDERules<MetricType, KernelType, TreeType>::
Score(const size_t queryIndex, TreeType& referenceNode)
{
  double score, maxKernel, minKernel, bound;
  const arma::vec& queryPoint = querySet.unsafe_col(queryIndex);
  const double minDistance = referenceNode.MinDistance(queryPoint);
  bool newCalculations = true;

  if (tree::TreeTraits<TreeType>::FirstPointIsCentroid &&
      lastQueryIndex == queryIndex &&
      traversalInfo.LastReferenceNode() != NULL &&
      traversalInfo.LastReferenceNode()->Point(0) == referenceNode.Point(0))
  {
    // Don't duplicate calculations.
    newCalculations = false;
    lastQueryIndex = queryIndex;
    lastReferenceIndex = referenceNode.Point(0);
  }
  else
  {
    // Calculations are new.
    maxKernel = kernel.Evaluate(minDistance);
    minKernel = kernel.Evaluate(referenceNode.MaxDistance(queryPoint));
    bound = maxKernel - minKernel;
  }

  if (newCalculations &&
      bound <= (absError + relError * minKernel) / referenceSet.n_cols)
  {
    // Estimate values.
    double kernelValue;

    // Calculate kernel value based on reference node centroid.
    if (tree::TreeTraits<TreeType>::FirstPointIsCentroid)
    {
      kernelValue = EvaluateKernel(queryIndex, referenceNode.Point(0));
    }
    else
    {
      kde::KDEStat& referenceStat = referenceNode.Stat();
      kernelValue = EvaluateKernel(queryPoint, referenceStat.Centroid());
    }

    densities(queryIndex) += referenceNode.NumDescendants() * kernelValue;

    // Don't explore this tree branch.
    score = DBL_MAX;
  }
  else
  {
    score = minDistance;
  }

  ++scores;
  traversalInfo.LastReferenceNode() = &referenceNode;
  traversalInfo.LastScore() = score;
  return score;
}

double RangeSearchRules<MetricType, TreeType>::Score(const size_t queryIndex,
                                                     TreeType& referenceNode)
{
  // We must get the minimum and maximum distances and store them in this
  // object.
  math::Range distances;

  if (tree::TreeTraits<TreeType>::FirstPointIsCentroid)
  {
    // In this situation, we calculate the base case.  So we should check to be
    // sure we haven't already done that.
    double baseCase;
    if (tree::TreeTraits<TreeType>::HasSelfChildren &&
        (referenceNode.Parent() != NULL) &&
        (referenceNode.Point(0) == referenceNode.Parent()->Point(0)))
    {
      // If the tree has self-children and this is a self-child, the base case
      // was already calculated.
      baseCase = referenceNode.Parent()->Stat().LastDistance();
      lastQueryIndex = queryIndex;
      lastReferenceIndex = referenceNode.Point(0);
    }
    else
    {
      // We must calculate the base case by hand.
      baseCase = BaseCase(queryIndex, referenceNode.Point(0));
    }

    // This may be possibly loose for non-ball bound trees.
    distances.Lo() = baseCase - referenceNode.FurthestDescendantDistance();
    distances.Hi() = baseCase + referenceNode.FurthestDescendantDistance();

    // Update last distance calculation.
    referenceNode.Stat().LastDistance() = baseCase;
  }
  else
  {
    distances = referenceNode.RangeDistance(querySet.unsafe_col(queryIndex));
  }

  // If the ranges do not overlap, prune this node.
  if (!distances.Contains(range))
    return DBL_MAX;

  // In this case, all of the points in the reference node will be part of the
  // results.
  if ((distances.Lo() >= range.Lo()) && (distances.Hi() <= range.Hi()))
  {
    AddResult(queryIndex, referenceNode);
    return DBL_MAX; // We don't need to go any deeper.
  }

  // Otherwise the score doesn't matter.  Recursion order is irrelevant in
  // range search.
  return 0.0;
}

 DTBStat(const TreeType& node) :
     maxNeighborDistance(DBL_MAX),
     minNeighborDistance(DBL_MAX),
     bound(DBL_MAX),
     componentMembership(
         ((node.NumPoints() == 1) && (node.NumChildren() == 0)) ?
           node.Point(0) : -1) { }

  DualTreeKMeansStatistic(TreeType& node) :
      neighbor::NeighborSearchStat<neighbor::NearestNeighborSort>(),
      upperBound(DBL_MAX),
      lowerBound(DBL_MAX),
      owner(size_t(-1)),
      pruned(size_t(-1)),
      staticPruned(false),
      staticUpperBoundMovement(0.0),
      staticLowerBoundMovement(0.0),
      trueParent(node.Parent())
  {
    // Empirically calculate the centroid.
    centroid.zeros(node.Dataset().n_rows);
    for (size_t i = 0; i < node.NumPoints(); ++i)
    {
      // Correct handling of cover tree: don't double-count the point which
      // appears in the children.
      if (tree::TreeTraits<TreeType>::HasSelfChildren && i == 0 &&
          node.NumChildren() > 0)
        continue;
      centroid += node.Dataset().col(node.Point(i));
    }

    for (size_t i = 0; i < node.NumChildren(); ++i)
      centroid += node.Child(i).NumDescendants() *
          node.Child(i).Stat().Centroid();

    centroid /= node.NumDescendants();

    // Set the true children correctly.
    trueChildren.resize(node.NumChildren());
    for (size_t i = 0; i < node.NumChildren(); ++i)
      trueChildren[i] = &node.Child(i);
  }

inline double NeighborSearchRules<SortPolicy, MetricType, TreeType>::Score(
    const size_t queryIndex,
    TreeType& referenceNode)
{
  ++scores; // Count number of Score() calls.
  double distance;
  if (tree::TreeTraits<TreeType>::FirstPointIsCentroid)
  {
    // The first point in the tree is the centroid.  So we can then calculate
    // the base case between that and the query point.
    double baseCase = -1.0;
    if (tree::TreeTraits<TreeType>::HasSelfChildren)
    {
      // If the parent node is the same, then we have already calculated the
      // base case.
      if ((referenceNode.Parent() != NULL) &&
          (referenceNode.Point(0) == referenceNode.Parent()->Point(0)))
        baseCase = referenceNode.Parent()->Stat().LastDistance();
      else
        baseCase = BaseCase(queryIndex, referenceNode.Point(0));

      // Save this evaluation.
      referenceNode.Stat().LastDistance() = baseCase;
    }

    distance = SortPolicy::CombineBest(baseCase,
        referenceNode.FurthestDescendantDistance());
  }
  else
  {
    distance = SortPolicy::BestPointToNodeDistance(querySet.col(queryIndex),
        &referenceNode);
  }

  // Compare against the best k'th distance for this query point so far.
  const double bestDistance = distances(distances.n_rows - 1, queryIndex);

  return (SortPolicy::IsBetter(distance, bestDistance)) ? distance : DBL_MAX;
}

inline double DTBRules<MetricType, TreeType>::CalculateBound(
    TreeType& queryNode) const
{
  double worstPointBound = -DBL_MAX;
  double bestPointBound = DBL_MAX;

  double worstChildBound = -DBL_MAX;
  double bestChildBound = DBL_MAX;

  // Now, find the best and worst point bounds.
  for (size_t i = 0; i < queryNode.NumPoints(); ++i)
  {
    const size_t pointComponent = connections.Find(queryNode.Point(i));
    const double bound = neighborsDistances[pointComponent];

    if (bound > worstPointBound)
      worstPointBound = bound;
    if (bound < bestPointBound)
      bestPointBound = bound;
  }

  // Find the best and worst child bounds.
  for (size_t i = 0; i < queryNode.NumChildren(); ++i)
  {
    const double maxBound = queryNode.Child(i).Stat().MaxNeighborDistance();
    if (maxBound > worstChildBound)
      worstChildBound = maxBound;

    const double minBound = queryNode.Child(i).Stat().MinNeighborDistance();
    if (minBound < bestChildBound)
      bestChildBound = minBound;
  }

  // Now calculate the actual bounds.
  const double worstBound = std::max(worstPointBound, worstChildBound);
  const double bestBound = std::min(bestPointBound, bestChildBound);
  // We must check that bestBound != DBL_MAX; otherwise, we risk overflow.
  const double bestAdjustedBound = (bestBound == DBL_MAX) ? DBL_MAX :
      bestBound + 2 * queryNode.FurthestDescendantDistance();

  // Update the relevant quantities in the node.
  queryNode.Stat().MaxNeighborDistance() = worstBound;
  queryNode.Stat().MinNeighborDistance() = bestBound;
  queryNode.Stat().Bound() = std::min(worstBound, bestAdjustedBound);

  return queryNode.Stat().Bound();
}

  DualTreeKMeansStatistic(TreeType& node) :
      closestQueryNode(NULL),
      minQueryNodeDistance(DBL_MAX),
      maxQueryNodeDistance(DBL_MAX),
      clustersPruned(0),
      iteration(size_t() - 1)
  {
    // Empirically calculate the centroid.
    centroid.zeros(node.Dataset().n_rows);
    for (size_t i = 0; i < node.NumPoints(); ++i)
      centroid += node.Dataset().col(node.Point(i));

    for (size_t i = 0; i < node.NumChildren(); ++i)
      centroid += node.Child(i).NumDescendants() *
          node.Child(i).Stat().Centroid();

    centroid /= node.NumDescendants();
  }

    PellegMooreKMeansStatistic(TreeType& node)
    {
        centroid.zeros(node.Dataset().n_rows);

        // Hope it's a depth-first build procedure.  Also, this won't work right for
        // trees that have self-children or stuff like that.
        for (size_t i = 0; i < node.NumChildren(); ++i)
        {
            centroid += node.Child(i).NumDescendants() *
                        node.Child(i).Stat().Centroid();
        }

        for (size_t i = 0; i < node.NumPoints(); ++i)
        {
            centroid += node.Dataset().col(node.Point(i));
        }

        if (node.NumDescendants() > 0)
            centroid /= node.NumDescendants();
        else
            centroid.fill(DBL_MAX); // Invalid centroid.  What else can we do?
    }

void GreedySingleTreeTraverser<TreeType, RuleType>::Traverse(
    const size_t queryIndex,
    TreeType& referenceNode)
{
  // Run the base case as necessary for all the points in the reference node.
  for (size_t i = 0; i < referenceNode.NumPoints(); ++i)
    rule.BaseCase(queryIndex, referenceNode.Point(i));

  size_t bestChild = rule.GetBestChild(queryIndex, referenceNode);
  size_t numDescendants;

  // Check that referencenode is not a leaf node while calculating number of
  // descendants of it's best child.
  if (!referenceNode.IsLeaf())
    numDescendants = referenceNode.Child(bestChild).NumDescendants();
  else
    numDescendants = referenceNode.NumPoints();

  // If number of descendants are more than minBaseCases than we can go along
  // with best child otherwise we need to traverse for each descendant to
  // ensure that we calculate at least minBaseCases number of base cases.
  if (!referenceNode.IsLeaf())
  {
    if (numDescendants > minBaseCases)
    {
      // We are prunning all but one child.
      numPrunes += referenceNode.NumChildren() - 1;
      // Recurse the best child.
      Traverse(queryIndex, referenceNode.Child(bestChild));
    }
    else
    {
      // Run the base case over first minBaseCases number of descendants.
      for (size_t i = 0; i <= minBaseCases; ++i)
        rule.BaseCase(queryIndex, referenceNode.Descendant(i));
    }
  }
}

void RangeSearchRules<MetricType, TreeType>::AddResult(const size_t queryIndex,
                                                       TreeType& referenceNode)
{
  // Some types of trees calculate the base case evaluation before Score() is
  // called, so if the base case has already been calculated, then we must avoid
  // adding that point to the results again.
  size_t baseCaseMod = 0;
  if (tree::TreeTraits<TreeType>::FirstPointIsCentroid &&
      (queryIndex == lastQueryIndex) &&
      (referenceNode.Point(0) == lastReferenceIndex))
  {
    baseCaseMod = 1;
  }

  // Resize distances and neighbors vectors appropriately.  We have to use
  // reserve() and not resize(), because we don't know if we will encounter the
  // case where the datasets and points are the same (and we skip in that case).
  const size_t oldSize = neighbors[queryIndex].size();
  neighbors[queryIndex].reserve(oldSize + referenceNode.NumDescendants() -
      baseCaseMod);
  distances[queryIndex].reserve(oldSize + referenceNode.NumDescendants() -
      baseCaseMod);

  for (size_t i = baseCaseMod; i < referenceNode.NumDescendants(); ++i)
  {
    if ((&referenceSet == &querySet) &&
        (queryIndex == referenceNode.Descendant(i)))
      continue;

    const double distance = metric.Evaluate(querySet.unsafe_col(queryIndex),
        referenceNode.Dataset().unsafe_col(referenceNode.Descendant(i)));

    neighbors[queryIndex].push_back(referenceNode.Descendant(i));
    distances[queryIndex].push_back(distance);
  }
}

void CheckTrees(TreeType& tree,
                TreeType& xmlTree,
                TreeType& textTree,
                TreeType& binaryTree)
{
  const typename TreeType::Mat* dataset = &tree.Dataset();

  // Make sure that the data matrices are the same.
  if (tree.Parent() == NULL)
  {
    CheckMatrices(*dataset,
                  xmlTree.Dataset(),
                  textTree.Dataset(),
                  binaryTree.Dataset());

    // Also ensure that the other parents are null too.
    BOOST_REQUIRE_EQUAL(xmlTree.Parent(), (TreeType*) NULL);
    BOOST_REQUIRE_EQUAL(textTree.Parent(), (TreeType*) NULL);
    BOOST_REQUIRE_EQUAL(binaryTree.Parent(), (TreeType*) NULL);
  }

  // Make sure the number of children is the same.
  BOOST_REQUIRE_EQUAL(tree.NumChildren(), xmlTree.NumChildren());
  BOOST_REQUIRE_EQUAL(tree.NumChildren(), textTree.NumChildren());
  BOOST_REQUIRE_EQUAL(tree.NumChildren(), binaryTree.NumChildren());

  // Make sure the number of descendants is the same.
  BOOST_REQUIRE_EQUAL(tree.NumDescendants(), xmlTree.NumDescendants());
  BOOST_REQUIRE_EQUAL(tree.NumDescendants(), textTree.NumDescendants());
  BOOST_REQUIRE_EQUAL(tree.NumDescendants(), binaryTree.NumDescendants());

  // Make sure the number of points is the same.
  BOOST_REQUIRE_EQUAL(tree.NumPoints(), xmlTree.NumPoints());
  BOOST_REQUIRE_EQUAL(tree.NumPoints(), textTree.NumPoints());
  BOOST_REQUIRE_EQUAL(tree.NumPoints(), binaryTree.NumPoints());

  // Check that each point is the same.
  for (size_t i = 0; i < tree.NumPoints(); ++i)
  {
    BOOST_REQUIRE_EQUAL(tree.Point(i), xmlTree.Point(i));
    BOOST_REQUIRE_EQUAL(tree.Point(i), textTree.Point(i));
    BOOST_REQUIRE_EQUAL(tree.Point(i), binaryTree.Point(i));
  }

  // Check that the parent distance is the same.
  BOOST_REQUIRE_CLOSE(tree.ParentDistance(), xmlTree.ParentDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.ParentDistance(), textTree.ParentDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.ParentDistance(), binaryTree.ParentDistance(), 1e-8);

  // Check that the furthest descendant distance is the same.
  BOOST_REQUIRE_CLOSE(tree.FurthestDescendantDistance(),
      xmlTree.FurthestDescendantDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.FurthestDescendantDistance(),
      textTree.FurthestDescendantDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.FurthestDescendantDistance(),
      binaryTree.FurthestDescendantDistance(), 1e-8);

  // Check that the minimum bound distance is the same.
  BOOST_REQUIRE_CLOSE(tree.MinimumBoundDistance(),
      xmlTree.MinimumBoundDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.MinimumBoundDistance(),
      textTree.MinimumBoundDistance(), 1e-8);
  BOOST_REQUIRE_CLOSE(tree.MinimumBoundDistance(),
      binaryTree.MinimumBoundDistance(), 1e-8);

  // Recurse into the children.
  for (size_t i = 0; i < tree.NumChildren(); ++i)
  {
    // Check that the child dataset is the same.
    BOOST_REQUIRE_EQUAL(&xmlTree.Dataset(), &xmlTree.Child(i).Dataset());
    BOOST_REQUIRE_EQUAL(&textTree.Dataset(), &textTree.Child(i).Dataset());
    BOOST_REQUIRE_EQUAL(&binaryTree.Dataset(), &binaryTree.Child(i).Dataset());

    // Make sure the parent link is right.
    BOOST_REQUIRE_EQUAL(xmlTree.Child(i).Parent(), &xmlTree);
    BOOST_REQUIRE_EQUAL(textTree.Child(i).Parent(), &textTree);
    BOOST_REQUIRE_EQUAL(binaryTree.Child(i).Parent(), &binaryTree);

    CheckTrees(tree.Child(i), xmlTree.Child(i), textTree.Child(i),
        binaryTree.Child(i));
  }
}

//.........这里部分代码省略.........
    else
    {
        // The reference parent could be NULL, which does weird things and we have
        // to consider.
        if (traversalInfo.LastQueryNode() != NULL)
        {
            adjustedScore += refDescDist *
                             traversalInfo.LastQueryNode()->Stat().SelfKernel();
            dualRefTerm = refDescDist;
        }
        else
        {
            // This makes it so a child-parent (or parent-parent) prune is not
            // possible.
            dualRefTerm = 0.0;
            adjustedScore = bestKernel;
        }
    }

    // Now add the dual term.
    adjustedScore += (dualQueryTerm * dualRefTerm);

    if (adjustedScore < bestKernel)
    {
        // It is not possible that this node combination can contain a point
        // combination with kernel value better than the minimum kernel value to
        // improve any of the results, so we can prune it.
        return DBL_MAX;
    }

    // We were unable to perform a parent-child or parent-parent prune, so now we
    // must calculate kernel evaluation, if necessary.
    double kernelEval = 0.0;
    if (tree::TreeTraits<TreeType>::FirstPointIsCentroid)
    {
        // For this type of tree, we may have already calculated the base case in
        // the parents.
        if ((traversalInfo.LastQueryNode() != NULL) &&
                (traversalInfo.LastReferenceNode() != NULL) &&
                (traversalInfo.LastQueryNode()->Point(0) == queryNode.Point(0)) &&
                (traversalInfo.LastReferenceNode()->Point(0) == referenceNode.Point(0)))
        {
            // Base case already done.
            kernelEval = traversalInfo.LastBaseCase();

            // When BaseCase() is called after Score(), these must be correct so that
            // another kernel evaluation is not performed.
            lastQueryIndex = queryNode.Point(0);
            lastReferenceIndex = referenceNode.Point(0);
        }
        else
        {
            // The kernel must be evaluated, but it is between points in the dataset,
            // so we can call BaseCase().  BaseCase() will set lastQueryIndex and
            // lastReferenceIndex correctly.
            kernelEval = BaseCase(queryNode.Point(0), referenceNode.Point(0));
        }

        traversalInfo.LastBaseCase() = kernelEval;
    }
    else
    {
        // Calculate the maximum possible kernel value.
        arma::vec queryCentroid;
        arma::vec refCentroid;
        queryNode.Centroid(queryCentroid);

inline double NeighborSearchRules<SortPolicy, MetricType, TreeType>::
    CalculateBound(TreeType& queryNode) const
{
  // We have five possible bounds, and we must take the best of them all.  We
  // don't use min/max here, but instead "best/worst", because this is general
  // to the nearest-neighbors/furthest-neighbors cases.  For nearest neighbors,
  // min = best, max = worst.
  //
  // (1) worst ( worst_{all points p in queryNode} D_p[k],
  //             worst_{all children c in queryNode} B(c) );
  // (2) best_{all points p in queryNode} D_p[k] + worst child distance +
  //        worst descendant distance;
  // (3) best_{all children c in queryNode} B(c) +
  //      2 ( worst descendant distance of queryNode -
  //          worst descendant distance of c );
  // (4) B_1(parent of queryNode)
  // (5) B_2(parent of queryNode);
  //
  // D_p[k] is the current k'th candidate distance for point p.
  // So we will loop over the points in queryNode and the children in queryNode
  // to calculate all five of these quantities.

  // Hm, can we populate our distances vector with estimates from the parent?
  // This is written specifically for the cover tree and assumes only one point
  // in a node.
//  if (queryNode.Parent() != NULL && queryNode.NumPoints() > 0)
//  {
//    size_t parentIndexStart = 0;
//    for (size_t i = 0; i < neighbors.n_rows; ++i)
//    {
//      const double pointDistance = distances(i, queryNode.Point(0));
//      if (pointDistance == DBL_MAX)
//      {
//      // Cool, can we take an estimate from the parent?
//        const double parentWorstBound = distances(distances.n_rows - 1,
//              queryNode.Parent()->Point(0));
//        if (parentWorstBound != DBL_MAX)
//        {
//          const double parentAdjustedDistance = parentWorstBound +
//              queryNode.ParentDistance();
//          distances(i, queryNode.Point(0)) = parentAdjustedDistance;
//        }
//      }
//    }
//  }

  double worstPointDistance = SortPolicy::BestDistance();
  double bestPointDistance = SortPolicy::WorstDistance();

  // Loop over all points in this node to find the best and worst distance
  // candidates (for (1) and (2)).
  for (size_t i = 0; i < queryNode.NumPoints(); ++i)
  {
    const double distance = distances(distances.n_rows - 1,
        queryNode.Point(i));
    if (SortPolicy::IsBetter(distance, bestPointDistance))
      bestPointDistance = distance;
    if (SortPolicy::IsBetter(worstPointDistance, distance))
      worstPointDistance = distance;
  }

  // Loop over all the children in this node to find the worst bound (for (1))
  // and the best bound with the correcting factor for descendant distances (for
  // (3)).
  double worstChildBound = SortPolicy::BestDistance();
  double bestAdjustedChildBound = SortPolicy::WorstDistance();
  const double queryMaxDescendantDistance =
      queryNode.FurthestDescendantDistance();

  for (size_t i = 0; i < queryNode.NumChildren(); ++i)
  {
    const double firstBound = queryNode.Child(i).Stat().FirstBound();
    const double secondBound = queryNode.Child(i).Stat().SecondBound();
    const double childMaxDescendantDistance =
        queryNode.Child(i).FurthestDescendantDistance();

    if (SortPolicy::IsBetter(worstChildBound, firstBound))
      worstChildBound = firstBound;

    // Now calculate adjustment for maximum descendant distances.
    const double adjustedBound = SortPolicy::CombineWorst(secondBound,
        2 * (queryMaxDescendantDistance - childMaxDescendantDistance));
    if (SortPolicy::IsBetter(adjustedBound, bestAdjustedChildBound))
      bestAdjustedChildBound = adjustedBound;
  }

  // This is bound (1).
  const double firstBound =
      (SortPolicy::IsBetter(worstPointDistance, worstChildBound)) ?
      worstChildBound : worstPointDistance;

  // This is bound (2).
  const double secondBound = SortPolicy::CombineWorst(
      SortPolicy::CombineWorst(bestPointDistance, queryMaxDescendantDistance),
      queryNode.FurthestPointDistance());

  // Bound (3) is bestAdjustedChildBound.

  // Bounds (4) and (5) are the parent bounds.
  const double fourthBound = (queryNode.Parent() != NULL) ?
//.........这里部分代码省略.........