Merge pull request #1263 from borglab/feature/nonlinear-hybrid

Nonlinear Hybrid

gtsam/discrete/discrete.i

@@ -219,6 +219,16 @@ class DiscreteBayesTree {
 };
 
 #include <gtsam/discrete/DiscreteLookupDAG.h>
+
+class DiscreteLookupTable : gtsam::DiscreteConditional{
+  DiscreteLookupTable(size_t nFrontals, const gtsam::DiscreteKeys& keys,
+                      const gtsam::DecisionTreeFactor::ADT& potentials);
+  void print(
+    const std::string& s = "Discrete Lookup Table: ",
+    const gtsam::KeyFormatter& keyFormatter = gtsam::DefaultKeyFormatter) const;
+  size_t argmax(const gtsam::DiscreteValues& parentsValues) const;
+};
+
 class DiscreteLookupDAG {
   DiscreteLookupDAG();
   void push_back(const gtsam::DiscreteLookupTable* table);

gtsam/hybrid/GaussianMixtureFactor.cpp

@@ -51,7 +51,7 @@ GaussianMixtureFactor GaussianMixtureFactor::FromFactors(
 void GaussianMixtureFactor::print(const std::string &s,
                                   const KeyFormatter &formatter) const {
   HybridFactor::print(s, formatter);
-  std::cout << "]{\n";
+  std::cout << "{\n";
   factors_.print(
       "", [&](Key k) { return formatter(k); },
       [&](const GaussianFactor::shared_ptr &gf) -> std::string {

gtsam/hybrid/GaussianMixtureFactor.h

@@ -22,7 +22,7 @@
 
 #include <gtsam/discrete/DecisionTree.h>
 #include <gtsam/discrete/DiscreteKey.h>
-#include <gtsam/hybrid/HybridFactor.h>
+#include <gtsam/hybrid/HybridGaussianFactor.h>
 #include <gtsam/linear/GaussianFactor.h>
 
 namespace gtsam {

gtsam/hybrid/HybridBayesTree.h

@@ -76,7 +76,10 @@ class GTSAM_EXPORT HybridBayesTree : public BayesTree<HybridBayesTreeClique> {
 /**
  * @brief Class for Hybrid Bayes tree orphan subtrees.
  *
- * This does special stuff for the hybrid case
+ * This object stores parent keys in our base type factor so that
+ * eliminating those parent keys will pull this subtree into the
+ * elimination.
+ * This does special stuff for the hybrid case.
  *
  * @tparam CLIQUE
  */

gtsam/hybrid/HybridFactor.cpp

@@ -89,17 +89,23 @@ void HybridFactor::print(const std::string &s,
   if (isContinuous_) std::cout << "Continuous ";
   if (isDiscrete_) std::cout << "Discrete ";
   if (isHybrid_) std::cout << "Hybrid ";
-  for (size_t c=0; c<continuousKeys_.size(); c++) {
+  std::cout << "[";
+  for (size_t c = 0; c < continuousKeys_.size(); c++) {
     std::cout << formatter(continuousKeys_.at(c));
     if (c < continuousKeys_.size() - 1) {
       std::cout << " ";
     } else {
-      std::cout << "; ";
+      std::cout << (discreteKeys_.size() > 0 ? "; " : "");
     }
   }
-  for(auto && discreteKey: discreteKeys_) {
+  for (size_t d = 0; d < discreteKeys_.size(); d++) {
-    std::cout << formatter(discreteKey.first) << " ";
+    std::cout << formatter(discreteKeys_.at(d).first);
+    if (d < discreteKeys_.size() - 1) {
+      std::cout << " ";
+    }
+    
   }
+  std::cout << "]";
 }
 
 }  // namespace gtsam

gtsam/hybrid/HybridGaussianFactor.h

@@ -37,6 +37,8 @@ class GTSAM_EXPORT HybridGaussianFactor : public HybridFactor {
   using This = HybridGaussianFactor;
   using shared_ptr = boost::shared_ptr<This>;
 
+  HybridGaussianFactor() = default;
+
   // Explicit conversion from a shared ptr of GF
   explicit HybridGaussianFactor(GaussianFactor::shared_ptr other);
 
@@ -52,7 +54,7 @@ class GTSAM_EXPORT HybridGaussianFactor : public HybridFactor {
 
   /// GTSAM print utility.
   void print(
-      const std::string &s = "HybridFactor\n",
+      const std::string &s = "HybridGaussianFactor\n",
       const KeyFormatter &formatter = DefaultKeyFormatter) const override;
 
   /// @}

gtsam/hybrid/HybridGaussianFactorGraph.cpp

@@ -98,6 +98,12 @@ GaussianMixtureFactor::Sum sumFrontals(
     } else if (f->isContinuous()) {
       deferredFactors.push_back(
           boost::dynamic_pointer_cast<HybridGaussianFactor>(f)->inner());
+
+    } else if (f->isDiscrete()) {
+      // Don't do anything for discrete-only factors
+      // since we want to eliminate continuous values only.
+      continue;
+
     } else {
       // We need to handle the case where the object is actually an
       // BayesTreeOrphanWrapper!
@@ -106,8 +112,8 @@ GaussianMixtureFactor::Sum sumFrontals(
       if (!orphan) {
         auto &fr = *f;
         throw std::invalid_argument(
-            std::string("factor is discrete in continuous elimination") +
+            std::string("factor is discrete in continuous elimination ") +
-            typeid(fr).name());
+            demangle(typeid(fr).name()));
       }
     }
   }
@@ -158,7 +164,7 @@ discreteElimination(const HybridGaussianFactorGraph &factors,
     }
   }
 
-  auto result = EliminateDiscrete(dfg, frontalKeys);
+  auto result = EliminateForMPE(dfg, frontalKeys);
 
   return {boost::make_shared<HybridConditional>(result.first),
           boost::make_shared<HybridDiscreteFactor>(result.second)};

gtsam/hybrid/HybridGaussianFactorGraph.h

@@ -24,6 +24,7 @@
 #include <gtsam/inference/EliminateableFactorGraph.h>
 #include <gtsam/inference/FactorGraph.h>
 #include <gtsam/inference/Ordering.h>
+#include <gtsam/linear/GaussianFactor.h>
 
 namespace gtsam {
 

gtsam/hybrid/HybridNonlinearFactor.cpp

@@ -0,0 +1,41 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ *  @file HybridNonlinearFactor.cpp
+ *  @date May 28, 2022
+ *  @author Varun Agrawal
+ */
+
+#include <gtsam/hybrid/HybridNonlinearFactor.h>
+
+#include <boost/make_shared.hpp>
+
+namespace gtsam {
+
+/* ************************************************************************* */
+HybridNonlinearFactor::HybridNonlinearFactor(
+    const NonlinearFactor::shared_ptr &other)
+    : Base(other->keys()), inner_(other) {}
+
+/* ************************************************************************* */
+bool HybridNonlinearFactor::equals(const HybridFactor &lf, double tol) const {
+  return Base::equals(lf, tol);
+}
+
+/* ************************************************************************* */
+void HybridNonlinearFactor::print(const std::string &s,
+                                  const KeyFormatter &formatter) const {
+  HybridFactor::print(s, formatter);
+  inner_->print("\n", formatter);
+};
+
+}  // namespace gtsam

gtsam/hybrid/HybridNonlinearFactor.h

@@ -0,0 +1,62 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ *  @file HybridNonlinearFactor.h
+ *  @date May 28, 2022
+ *  @author Varun Agrawal
+ */
+
+#pragma once
+
+#include <gtsam/hybrid/HybridFactor.h>
+#include <gtsam/hybrid/HybridGaussianFactor.h>
+#include <gtsam/nonlinear/NonlinearFactor.h>
+
+namespace gtsam {
+
+/**
+ * A HybridNonlinearFactor is a layer over NonlinearFactor so that we do not
+ * have a diamond inheritance.
+ */
+class HybridNonlinearFactor : public HybridFactor {
+ private:
+  NonlinearFactor::shared_ptr inner_;
+
+ public:
+  using Base = HybridFactor;
+  using This = HybridNonlinearFactor;
+  using shared_ptr = boost::shared_ptr<This>;
+
+  // Explicit conversion from a shared ptr of NonlinearFactor
+  explicit HybridNonlinearFactor(const NonlinearFactor::shared_ptr &other);
+
+  /// @name Testable
+  /// @{
+
+  /// Check equality.
+  virtual bool equals(const HybridFactor &lf, double tol) const override;
+
+  /// GTSAM print utility.
+  void print(
+      const std::string &s = "HybridNonlinearFactor\n",
+      const KeyFormatter &formatter = DefaultKeyFormatter) const override;
+
+  /// @}
+
+  NonlinearFactor::shared_ptr inner() const { return inner_; }
+
+  /// Linearize to a HybridGaussianFactor at the linearization point `c`.
+  boost::shared_ptr<HybridGaussianFactor> linearize(const Values &c) const {
+    return boost::make_shared<HybridGaussianFactor>(inner_->linearize(c));
+  }
+};
+}  // namespace gtsam

gtsam/hybrid/HybridNonlinearFactorGraph.cpp

@@ -0,0 +1,94 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file   HybridNonlinearFactorGraph.cpp
+ * @brief  Nonlinear hybrid factor graph that uses type erasure
+ * @author Varun Agrawal
+ * @date   May 28, 2022
+ */
+
+#include <gtsam/hybrid/HybridNonlinearFactorGraph.h>
+
+namespace gtsam {
+
+/* ************************************************************************* */
+void HybridNonlinearFactorGraph::add(
+    boost::shared_ptr<NonlinearFactor> factor) {
+  FactorGraph::add(boost::make_shared<HybridNonlinearFactor>(factor));
+}
+
+/* ************************************************************************* */
+void HybridNonlinearFactorGraph::print(const std::string& s,
+                                       const KeyFormatter& keyFormatter) const {
+  // Base::print(str, keyFormatter);
+  std::cout << (s.empty() ? "" : s + " ") << std::endl;
+  std::cout << "size: " << size() << std::endl;
+  for (size_t i = 0; i < factors_.size(); i++) {
+    std::stringstream ss;
+    ss << "factor " << i << ": ";
+    if (factors_[i]) {
+      factors_[i]->print(ss.str(), keyFormatter);
+      std::cout << std::endl;
+    }
+  }
+}
+
+/* ************************************************************************* */
+HybridGaussianFactorGraph HybridNonlinearFactorGraph::linearize(
+    const Values& continuousValues) const {
+  // create an empty linear FG
+  HybridGaussianFactorGraph linearFG;
+
+  linearFG.reserve(size());
+
+  // linearize all hybrid factors
+  for (auto&& factor : factors_) {
+    // First check if it is a valid factor
+    if (factor) {
+      // Check if the factor is a hybrid factor.
+      // It can be either a nonlinear MixtureFactor or a linear
+      // GaussianMixtureFactor.
+      if (factor->isHybrid()) {
+        // Check if it is a nonlinear mixture factor
+        if (auto nlmf = boost::dynamic_pointer_cast<MixtureFactor>(factor)) {
+          linearFG.push_back(nlmf->linearize(continuousValues));
+        } else {
+          linearFG.push_back(factor);
+        }
+
+        // Now check if the factor is a continuous only factor.
+      } else if (factor->isContinuous()) {
+        // In this case, we check if factor->inner() is nonlinear since
+        // HybridFactors wrap over continuous factors.
+        auto nlhf = boost::dynamic_pointer_cast<HybridNonlinearFactor>(factor);
+        if (auto nlf =
+                boost::dynamic_pointer_cast<NonlinearFactor>(nlhf->inner())) {
+          auto hgf = boost::make_shared<HybridGaussianFactor>(
+              nlf->linearize(continuousValues));
+          linearFG.push_back(hgf);
+        } else {
+          linearFG.push_back(factor);
+        }
+        // Finally if nothing else, we are discrete-only which doesn't need
+        // lineariztion.
+      } else {
+        linearFG.push_back(factor);
+      }
+
+    } else {
+      linearFG.push_back(GaussianFactor::shared_ptr());
+    }
+  }
+  return linearFG;
+}
+
+}  // namespace gtsam

gtsam/hybrid/HybridNonlinearFactorGraph.h

@@ -0,0 +1,134 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file   HybridNonlinearFactorGraph.h
+ * @brief  Nonlinear hybrid factor graph that uses type erasure
+ * @author Varun Agrawal
+ * @date   May 28, 2022
+ */
+
+#pragma once
+
+#include <gtsam/hybrid/HybridFactor.h>
+#include <gtsam/hybrid/HybridFactorGraph.h>
+#include <gtsam/hybrid/HybridGaussianFactorGraph.h>
+#include <gtsam/hybrid/HybridNonlinearFactor.h>
+#include <gtsam/hybrid/MixtureFactor.h>
+#include <gtsam/inference/Ordering.h>
+#include <gtsam/nonlinear/NonlinearFactor.h>
+
+#include <boost/format.hpp>
+namespace gtsam {
+
+/**
+ * Nonlinear Hybrid Factor Graph
+ * -----------------------
+ * This is the non-linear version of a hybrid factor graph.
+ * Everything inside needs to be hybrid factor or hybrid conditional.
+ */
+class GTSAM_EXPORT HybridNonlinearFactorGraph : public HybridFactorGraph {
+ protected:
+  /// Check if FACTOR type is derived from NonlinearFactor.
+  template <typename FACTOR>
+  using IsNonlinear = typename std::enable_if<
+      std::is_base_of<NonlinearFactor, FACTOR>::value>::type;
+
+ public:
+  using Base = HybridFactorGraph;
+  using This = HybridNonlinearFactorGraph;     ///< this class
+  using shared_ptr = boost::shared_ptr<This>;  ///< shared_ptr to This
+
+  using Values = gtsam::Values;  ///< backwards compatibility
+  using Indices = KeyVector;     ///> map from keys to values
+
+  /// @name Constructors
+  /// @{
+
+  HybridNonlinearFactorGraph() = default;
+
+  /**
+   * Implicit copy/downcast constructor to override explicit template container
+   * constructor. In BayesTree this is used for:
+   * `cachedSeparatorMarginal_.reset(*separatorMarginal)`
+   * */
+  template <class DERIVEDFACTOR>
+  HybridNonlinearFactorGraph(const FactorGraph<DERIVEDFACTOR>& graph)
+      : Base(graph) {}
+
+  /// @}
+
+  // Allow use of selected FactorGraph methods:
+  using Base::empty;
+  using Base::reserve;
+  using Base::size;
+  using Base::operator[];
+  using Base::add;
+  using Base::resize;
+
+  /**
+   * Add a nonlinear factor *pointer* to the internal nonlinear factor graph
+   * @param nonlinearFactor - boost::shared_ptr to the factor to add
+   */
+  template <typename FACTOR>
+  IsNonlinear<FACTOR> push_nonlinear(
+      const boost::shared_ptr<FACTOR>& nonlinearFactor) {
+    Base::push_back(boost::make_shared<HybridNonlinearFactor>(nonlinearFactor));
+  }
+
+  /// Construct a factor and add (shared pointer to it) to factor graph.
+  template <class FACTOR, class... Args>
+  IsNonlinear<FACTOR> emplace_nonlinear(Args&&... args) {
+    auto factor = boost::allocate_shared<FACTOR>(
+        Eigen::aligned_allocator<FACTOR>(), std::forward<Args>(args)...);
+    push_nonlinear(factor);
+  }
+
+  /**
+   * @brief Add a single factor shared pointer to the hybrid factor graph.
+   * Dynamically handles the factor type and assigns it to the correct
+   * underlying container.
+   *
+   * @tparam FACTOR The factor type template
+   * @param sharedFactor The factor to add to this factor graph.
+   */
+  template <typename FACTOR>
+  void push_back(const boost::shared_ptr<FACTOR>& sharedFactor) {
+    if (auto p = boost::dynamic_pointer_cast<NonlinearFactor>(sharedFactor)) {
+      push_nonlinear(p);
+    } else {
+      Base::push_back(sharedFactor);
+    }
+  }
+
+  /// Add a nonlinear factor as a shared ptr.
+  void add(boost::shared_ptr<NonlinearFactor> factor);
+
+  /// Print the factor graph.
+  void print(
+      const std::string& s = "HybridNonlinearFactorGraph",
+      const KeyFormatter& keyFormatter = DefaultKeyFormatter) const override;
+
+  /**
+   * @brief Linearize all the continuous factors in the
+   * HybridNonlinearFactorGraph.
+   *
+   * @param continuousValues: Dictionary of continuous values.
+   * @return HybridGaussianFactorGraph::shared_ptr
+   */
+  HybridGaussianFactorGraph linearize(const Values& continuousValues) const;
+};
+
+template <>
+struct traits<HybridNonlinearFactorGraph>
+    : public Testable<HybridNonlinearFactorGraph> {};
+
+}  // namespace gtsam

gtsam/hybrid/MixtureFactor.h

@@ -0,0 +1,244 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file   MixtureFactor.h
+ * @brief  Nonlinear Mixture factor of continuous and discrete.
+ * @author Kevin Doherty, kdoherty@mit.edu
+ * @author Varun Agrawal
+ * @date   December 2021
+ */
+
+#pragma once
+
+#include <gtsam/discrete/DiscreteValues.h>
+#include <gtsam/hybrid/GaussianMixtureFactor.h>
+#include <gtsam/hybrid/HybridNonlinearFactor.h>
+#include <gtsam/nonlinear/NonlinearFactor.h>
+#include <gtsam/nonlinear/Symbol.h>
+
+#include <algorithm>
+#include <boost/format.hpp>
+#include <cmath>
+#include <limits>
+#include <vector>
+
+namespace gtsam {
+
+/**
+ * @brief Implementation of a discrete conditional mixture factor.
+ *
+ * Implements a joint discrete-continuous factor where the discrete variable
+ * serves to "select" a mixture component corresponding to a NonlinearFactor
+ * type of measurement.
+ *
+ * This class stores all factors as HybridFactors which can then be typecast to
+ * one of (NonlinearFactor, GaussianFactor) which can then be checked to perform
+ * the correct operation.
+ */
+class MixtureFactor : public HybridFactor {
+ public:
+  using Base = HybridFactor;
+  using This = MixtureFactor;
+  using shared_ptr = boost::shared_ptr<MixtureFactor>;
+  using sharedFactor = boost::shared_ptr<NonlinearFactor>;
+
+  /**
+   * @brief typedef for DecisionTree which has Keys as node labels and
+   * NonlinearFactor as leaf nodes.
+   */
+  using Factors = DecisionTree<Key, sharedFactor>;
+
+ private:
+  /// Decision tree of Gaussian factors indexed by discrete keys.
+  Factors factors_;
+  bool normalized_;
+
+ public:
+  MixtureFactor() = default;
+
+  /**
+   * @brief Construct from Decision tree.
+   *
+   * @param keys Vector of keys for continuous factors.
+   * @param discreteKeys Vector of discrete keys.
+   * @param factors Decision tree with of shared factors.
+   * @param normalized Flag indicating if the factor error is already
+   * normalized.
+   */
+  MixtureFactor(const KeyVector& keys, const DiscreteKeys& discreteKeys,
+                const Factors& factors, bool normalized = false)
+      : Base(keys, discreteKeys), factors_(factors), normalized_(normalized) {}
+
+  /**
+   * @brief Convenience constructor that generates the underlying factor
+   * decision tree for us.
+   *
+   * Here it is important that the vector of factors has the correct number of
+   * elements based on the number of discrete keys and the cardinality of the
+   * keys, so that the decision tree is constructed appropriately.
+   *
+   * @tparam FACTOR The type of the factor shared pointers being passed in. Will
+   * be typecast to NonlinearFactor shared pointers.
+   * @param keys Vector of keys for continuous factors.
+   * @param discreteKeys Vector of discrete keys.
+   * @param factors Vector of shared pointers to factors.
+   * @param normalized Flag indicating if the factor error is already
+   * normalized.
+   */
+  template <typename FACTOR>
+  MixtureFactor(const KeyVector& keys, const DiscreteKeys& discreteKeys,
+                const std::vector<boost::shared_ptr<FACTOR>>& factors,
+                bool normalized = false)
+      : Base(keys, discreteKeys), normalized_(normalized) {
+    std::vector<NonlinearFactor::shared_ptr> nonlinear_factors;
+    for (auto&& f : factors) {
+      nonlinear_factors.push_back(
+          boost::dynamic_pointer_cast<NonlinearFactor>(f));
+    }
+    factors_ = Factors(discreteKeys, nonlinear_factors);
+  }
+
+  ~MixtureFactor() = default;
+
+  /**
+   * @brief Compute error of factor given both continuous and discrete values.
+   *
+   * @param continuousVals The continuous Values.
+   * @param discreteVals The discrete Values.
+   * @return double The error of this factor.
+   */
+  double error(const Values& continuousVals,
+               const DiscreteValues& discreteVals) const {
+    // Retrieve the factor corresponding to the assignment in discreteVals.
+    auto factor = factors_(discreteVals);
+    // Compute the error for the selected factor
+    const double factorError = factor->error(continuousVals);
+    if (normalized_) return factorError;
+    return factorError +
+           this->nonlinearFactorLogNormalizingConstant(factor, continuousVals);
+  }
+
+  size_t dim() const {
+    // TODO(Varun)
+    throw std::runtime_error("MixtureFactor::dim not implemented.");
+  }
+
+  /// Testable
+  /// @{
+
+  /// print to stdout
+  void print(
+      const std::string& s = "MixtureFactor",
+      const KeyFormatter& keyFormatter = DefaultKeyFormatter) const override {
+    std::cout << (s.empty() ? "" : s + " ");
+    Base::print("", keyFormatter);
+    std::cout << "\nMixtureFactor\n";
+    auto valueFormatter = [](const sharedFactor& v) {
+      if (v) {
+        return (boost::format("Nonlinear factor on %d keys") % v->size()).str();
+      } else {
+        return std::string("nullptr");
+      }
+    };
+    factors_.print("", keyFormatter, valueFormatter);
+  }
+
+  /// Check equality
+  bool equals(const HybridFactor& other, double tol = 1e-9) const override {
+    // We attempt a dynamic cast from HybridFactor to MixtureFactor. If it
+    // fails, return false.
+    if (!dynamic_cast<const MixtureFactor*>(&other)) return false;
+
+    // If the cast is successful, we'll properly construct a MixtureFactor
+    // object from `other`
+    const MixtureFactor& f(static_cast<const MixtureFactor&>(other));
+
+    // Ensure that this MixtureFactor and `f` have the same `factors_`.
+    auto compare = [tol](const sharedFactor& a, const sharedFactor& b) {
+      return traits<NonlinearFactor>::Equals(*a, *b, tol);
+    };
+    if (!factors_.equals(f.factors_, compare)) return false;
+
+    // If everything above passes, and the keys_, discreteKeys_ and normalized_
+    // member variables are identical, return true.
+    return (std::equal(keys_.begin(), keys_.end(), f.keys().begin()) &&
+            (discreteKeys_ == f.discreteKeys_) &&
+            (normalized_ == f.normalized_));
+  }
+
+  /// @}
+
+  /// Linearize specific nonlinear factors based on the assignment in
+  /// discreteValues.
+  GaussianFactor::shared_ptr linearize(
+      const Values& continuousVals, const DiscreteValues& discreteVals) const {
+    auto factor = factors_(discreteVals);
+    return factor->linearize(continuousVals);
+  }
+
+  /// Linearize all the continuous factors to get a GaussianMixtureFactor.
+  boost::shared_ptr<GaussianMixtureFactor> linearize(
+      const Values& continuousVals) const {
+    // functional to linearize each factor in the decision tree
+    auto linearizeDT = [continuousVals](const sharedFactor& factor) {
+      return factor->linearize(continuousVals);
+    };
+
+    DecisionTree<Key, GaussianFactor::shared_ptr> linearized_factors(
+        factors_, linearizeDT);
+
+    return boost::make_shared<GaussianMixtureFactor>(
+        continuousKeys_, discreteKeys_, linearized_factors);
+  }
+
+  /**
+   * If the component factors are not already normalized, we want to compute
+   * their normalizing constants so that the resulting joint distribution is
+   * appropriately computed. Remember, this is the _negative_ normalizing
+   * constant for the measurement likelihood (since we are minimizing the
+   * _negative_ log-likelihood).
+   */
+  double nonlinearFactorLogNormalizingConstant(const sharedFactor& factor,
+                                               const Values& values) const {
+    // Information matrix (inverse covariance matrix) for the factor.
+    Matrix infoMat;
+
+    // If this is a NoiseModelFactor, we'll use its noiseModel to
+    // otherwise noiseModelFactor will be nullptr
+    if (auto noiseModelFactor =
+            boost::dynamic_pointer_cast<NoiseModelFactor>(factor)) {
+      // If dynamic cast to NoiseModelFactor succeeded, see if the noise model
+      // is Gaussian
+      auto noiseModel = noiseModelFactor->noiseModel();
+
+      auto gaussianNoiseModel =
+          boost::dynamic_pointer_cast<noiseModel::Gaussian>(noiseModel);
+      if (gaussianNoiseModel) {
+        // If the noise model is Gaussian, retrieve the information matrix
+        infoMat = gaussianNoiseModel->information();
+      } else {
+        // If the factor is not a Gaussian factor, we'll linearize it to get
+        // something with a normalized noise model
+        // TODO(kevin): does this make sense to do? I think maybe not in
+        // general? Should we just yell at the user?
+        auto gaussianFactor = factor->linearize(values);
+        infoMat = gaussianFactor->information();
+      }
+    }
+
+    // Compute the (negative) log of the normalizing constant
+    return -(factor->dim() * log(2.0 * M_PI) / 2.0) -
+           (log(infoMat.determinant()) / 2.0);
+  }
+};
+
+}  // namespace gtsam

gtsam/hybrid/tests/Switching.h

@@ -21,10 +21,16 @@
 #include <gtsam/discrete/DiscreteDistribution.h>
 #include <gtsam/hybrid/GaussianMixtureFactor.h>
 #include <gtsam/hybrid/HybridGaussianFactorGraph.h>
+#include <gtsam/hybrid/HybridNonlinearFactorGraph.h>
+#include <gtsam/hybrid/MixtureFactor.h>
 #include <gtsam/inference/Symbol.h>
 #include <gtsam/linear/JacobianFactor.h>
+#include <gtsam/linear/NoiseModel.h>
+#include <gtsam/nonlinear/PriorFactor.h>
 #include <gtsam/slam/BetweenFactor.h>
 
+#include <vector>
+
 #pragma once
 
 namespace gtsam {
@@ -33,111 +39,6 @@ using symbol_shorthand::C;
 using symbol_shorthand::M;
 using symbol_shorthand::X;
 
-/* ****************************************************************************/
-// Test fixture with switching network.
-// TODO(Varun) Currently this is only linear. We need to add nonlinear support
-// and then update to
-// https://github.com/borglab/gtsam/pull/973/files#diff-58c02b3b197ebf731694946e87762d252e9eaa2f5c6c4ba22d618085b321ca23
-struct Switching {
-  size_t K;
-  DiscreteKeys modes;
-  HybridGaussianFactorGraph linearizedFactorGraph;
-  Values linearizationPoint;
-
-  using MotionModel = BetweenFactor<double>;
-  // using MotionMixture = MixtureFactor<MotionModel>;
-
-  /// Create with given number of time steps.
-  Switching(size_t K, double between_sigma = 1.0, double prior_sigma = 0.1)
-      : K(K) {
-    // Create DiscreteKeys for binary K modes, modes[0] will not be used.
-    modes = addDiscreteModes(K);
-
-    // Create hybrid factor graph.
-    // Add a prior on X(1).
-    auto prior = boost::make_shared<JacobianFactor>(
-        X(1), Matrix11::Ones() / prior_sigma, Vector1::Zero());
-    linearizedFactorGraph.push_back(prior);
-
-    // Add "motion models".
-    linearizedFactorGraph = addMotionModels(K);
-
-    // Add measurement factors
-    for (size_t k = 1; k <= K; k++) {
-      linearizedFactorGraph.emplace_gaussian<JacobianFactor>(
-          X(k), Matrix11::Ones() / 0.1, Vector1::Zero());
-    }
-
-    // Add "mode chain"
-    linearizedFactorGraph = addModeChain(linearizedFactorGraph);
-
-    // Create the linearization point.
-    for (size_t k = 1; k <= K; k++) {
-      linearizationPoint.insert<double>(X(k), static_cast<double>(k));
-    }
-  }
-
-  /// Create DiscreteKeys for K binary modes.
-  DiscreteKeys addDiscreteModes(size_t K) {
-    DiscreteKeys m;
-    for (size_t k = 0; k <= K; k++) {
-      m.emplace_back(M(k), 2);
-    }
-    return m;
-  }
-
-  /// Helper function to add motion models for each [k, k+1] interval.
-  HybridGaussianFactorGraph addMotionModels(size_t K) {
-    HybridGaussianFactorGraph hgfg;
-    for (size_t k = 1; k < K; k++) {
-      auto keys = {X(k), X(k + 1)};
-      auto components = motionModels(k);
-      hgfg.emplace_hybrid<GaussianMixtureFactor>(keys, DiscreteKeys{modes[k]},
-                                                 components);
-    }
-    return hgfg;
-  }
-
-  // Create motion models for a given time step
-  static std::vector<GaussianFactor::shared_ptr> motionModels(
-      size_t k, double sigma = 1.0) {
-    auto noise_model = noiseModel::Isotropic::Sigma(1, sigma);
-    auto still = boost::make_shared<JacobianFactor>(
-             X(k), -Matrix11::Ones() / sigma, X(k + 1),
-             Matrix11::Ones() / sigma, Vector1::Zero()),
-         moving = boost::make_shared<JacobianFactor>(
-             X(k), -Matrix11::Ones() / sigma, X(k + 1),
-             Matrix11::Ones() / sigma, -Vector1::Ones() / sigma);
-    return {boost::dynamic_pointer_cast<GaussianFactor>(still),
-            boost::dynamic_pointer_cast<GaussianFactor>(moving)};
-  }
-
-  // // Add "mode chain" to NonlinearHybridFactorGraph
-  // void addModeChain(HybridNonlinearFactorGraph& fg) {
-  //   auto prior = boost::make_shared<DiscreteDistribution>(modes[1], "1/1");
-  //   fg.push_discrete(prior);
-  //   for (size_t k = 1; k < K - 1; k++) {
-  //     auto parents = {modes[k]};
-  //     auto conditional = boost::make_shared<DiscreteConditional>(
-  //         modes[k + 1], parents, "1/2 3/2");
-  //     fg.push_discrete(conditional);
-  //   }
-  // }
-
-  // Add "mode chain" to GaussianHybridFactorGraph
-  HybridGaussianFactorGraph addModeChain(HybridGaussianFactorGraph& fg) {
-    auto prior = boost::make_shared<DiscreteDistribution>(modes[1], "1/1");
-    fg.push_discrete(prior);
-    for (size_t k = 1; k < K - 1; k++) {
-      auto parents = {modes[k]};
-      auto conditional = boost::make_shared<DiscreteConditional>(
-          modes[k + 1], parents, "1/2 3/2");
-      fg.push_discrete(conditional);
-    }
-    return fg;
-  }
-};
-
 /**
  * @brief Create a switching system chain. A switching system is a continuous
  * system which depends on a discrete mode at each time step of the chain.
@@ -149,7 +50,7 @@ struct Switching {
  */
 inline HybridGaussianFactorGraph::shared_ptr makeSwitchingChain(
     size_t n, std::function<Key(int)> keyFunc = X,
-    std::function<Key(int)> dKeyFunc = C) {
+    std::function<Key(int)> dKeyFunc = M) {
   HybridGaussianFactorGraph hfg;
 
   hfg.add(JacobianFactor(keyFunc(1), I_3x3, Z_3x1));
@@ -182,7 +83,7 @@ inline HybridGaussianFactorGraph::shared_ptr makeSwitchingChain(
  * @return std::pair<KeyVector, std::vector<int>>
  */
 inline std::pair<KeyVector, std::vector<int>> makeBinaryOrdering(
-    std::vector<Key>& input) {
+    std::vector<Key> &input) {
   KeyVector new_order;
 
   std::vector<int> levels(input.size());
@@ -211,4 +112,103 @@ inline std::pair<KeyVector, std::vector<int>> makeBinaryOrdering(
   return {new_order, levels};
 }
 
+/* ***************************************************************************
+ */
+using MotionModel = BetweenFactor<double>;
+// using MotionMixture = MixtureFactor<MotionModel>;
+
+// Test fixture with switching network.
+struct Switching {
+  size_t K;
+  DiscreteKeys modes;
+  HybridNonlinearFactorGraph nonlinearFactorGraph;
+  HybridGaussianFactorGraph linearizedFactorGraph;
+  Values linearizationPoint;
+
+  /// Create with given number of time steps.
+  Switching(size_t K, double between_sigma = 1.0, double prior_sigma = 0.1)
+      : K(K) {
+    using symbol_shorthand::M;
+    using symbol_shorthand::X;
+
+    // Create DiscreteKeys for binary K modes, modes[0] will not be used.
+    for (size_t k = 0; k <= K; k++) {
+      modes.emplace_back(M(k), 2);
+    }
+
+    // Create hybrid factor graph.
+    // Add a prior on X(1).
+    auto prior = boost::make_shared<PriorFactor<double>>(
+        X(1), 0, noiseModel::Isotropic::Sigma(1, prior_sigma));
+    nonlinearFactorGraph.push_nonlinear(prior);
+
+    // Add "motion models".
+    for (size_t k = 1; k < K; k++) {
+      KeyVector keys = {X(k), X(k + 1)};
+      auto motion_models = motionModels(k);
+      std::vector<NonlinearFactor::shared_ptr> components;
+      for (auto &&f : motion_models) {
+        components.push_back(boost::dynamic_pointer_cast<NonlinearFactor>(f));
+      }
+      nonlinearFactorGraph.emplace_hybrid<MixtureFactor>(
+          keys, DiscreteKeys{modes[k]}, components);
+    }
+
+    // Add measurement factors
+    auto measurement_noise = noiseModel::Isotropic::Sigma(1, 0.1);
+    for (size_t k = 2; k <= K; k++) {
+      nonlinearFactorGraph.emplace_nonlinear<PriorFactor<double>>(
+          X(k), 1.0 * (k - 1), measurement_noise);
+    }
+
+    // Add "mode chain"
+    addModeChain(&nonlinearFactorGraph);
+
+    // Create the linearization point.
+    for (size_t k = 1; k <= K; k++) {
+      linearizationPoint.insert<double>(X(k), static_cast<double>(k));
+    }
+
+    linearizedFactorGraph = nonlinearFactorGraph.linearize(linearizationPoint);
+  }
+
+  // Create motion models for a given time step
+  static std::vector<MotionModel::shared_ptr> motionModels(size_t k,
+                                                           double sigma = 1.0) {
+    using symbol_shorthand::M;
+    using symbol_shorthand::X;
+
+    auto noise_model = noiseModel::Isotropic::Sigma(1, sigma);
+    auto still =
+             boost::make_shared<MotionModel>(X(k), X(k + 1), 0.0, noise_model),
+         moving =
+             boost::make_shared<MotionModel>(X(k), X(k + 1), 1.0, noise_model);
+    return {still, moving};
+  }
+
+  // Add "mode chain" to HybridNonlinearFactorGraph
+  void addModeChain(HybridNonlinearFactorGraph *fg) {
+    auto prior = boost::make_shared<DiscreteDistribution>(modes[1], "1/1");
+    fg->push_discrete(prior);
+    for (size_t k = 1; k < K - 1; k++) {
+      auto parents = {modes[k]};
+      auto conditional = boost::make_shared<DiscreteConditional>(
+          modes[k + 1], parents, "1/2 3/2");
+      fg->push_discrete(conditional);
+    }
+  }
+
+  // Add "mode chain" to HybridGaussianFactorGraph
+  void addModeChain(HybridGaussianFactorGraph *fg) {
+    auto prior = boost::make_shared<DiscreteDistribution>(modes[1], "1/1");
+    fg->push_discrete(prior);
+    for (size_t k = 1; k < K - 1; k++) {
+      auto parents = {modes[k]};
+      auto conditional = boost::make_shared<DiscreteConditional>(
+          modes[k + 1], parents, "1/2 3/2");
+      fg->push_discrete(conditional);
+    }
+  }
+};
+
 }  // namespace gtsam

gtsam/hybrid/tests/testGaussianHybridFactorGraph.cpp

@@ -52,8 +52,8 @@ using namespace boost::assign;
 using namespace std;
 using namespace gtsam;
 
-using gtsam::symbol_shorthand::C;
 using gtsam::symbol_shorthand::D;
+using gtsam::symbol_shorthand::M;
 using gtsam::symbol_shorthand::X;
 using gtsam::symbol_shorthand::Y;
 
@@ -67,9 +67,9 @@ TEST(HybridGaussianFactorGraph, Creation) {
 
   // Define a gaussian mixture conditional P(x0|x1, c0) and add it to the factor
   // graph
-  GaussianMixture gm({X(0)}, {X(1)}, DiscreteKeys(DiscreteKey{C(0), 2}),
+  GaussianMixture gm({X(0)}, {X(1)}, DiscreteKeys(DiscreteKey{M(0), 2}),
                      GaussianMixture::Conditionals(
-                         C(0),
+                         M(0),
                          boost::make_shared<GaussianConditional>(
                              X(0), Z_3x1, I_3x3, X(1), I_3x3),
                          boost::make_shared<GaussianConditional>(
@@ -96,11 +96,11 @@ TEST(HybridGaussianFactorGraph, EliminateMultifrontal) {
   // Test multifrontal elimination
   HybridGaussianFactorGraph hfg;
 
-  DiscreteKey c(C(1), 2);
+  DiscreteKey m(M(1), 2);
 
   // Add priors on x0 and c1
   hfg.add(JacobianFactor(X(0), I_3x3, Z_3x1));
-  hfg.add(HybridDiscreteFactor(DecisionTreeFactor(c, {2, 8})));
+  hfg.add(HybridDiscreteFactor(DecisionTreeFactor(m, {2, 8})));
 
   Ordering ordering;
   ordering.push_back(X(0));
@@ -114,7 +114,7 @@ TEST(HybridGaussianFactorGraph, EliminateMultifrontal) {
 TEST(HybridGaussianFactorGraph, eliminateFullSequentialEqualChance) {
   HybridGaussianFactorGraph hfg;
 
-  DiscreteKey c1(C(1), 2);
+  DiscreteKey m1(M(1), 2);
 
   // Add prior on x0
   hfg.add(JacobianFactor(X(0), I_3x3, Z_3x1));
@@ -123,45 +123,45 @@ TEST(HybridGaussianFactorGraph, eliminateFullSequentialEqualChance) {
 
   // Add a gaussian mixture factor ϕ(x1, c1)
   DecisionTree<Key, GaussianFactor::shared_ptr> dt(
-      C(1), boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
+      M(1), boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
       boost::make_shared<JacobianFactor>(X(1), I_3x3, Vector3::Ones()));
 
-  hfg.add(GaussianMixtureFactor({X(1)}, {c1}, dt));
+  hfg.add(GaussianMixtureFactor({X(1)}, {m1}, dt));
 
   auto result =
-      hfg.eliminateSequential(Ordering::ColamdConstrainedLast(hfg, {C(1)}));
+      hfg.eliminateSequential(Ordering::ColamdConstrainedLast(hfg, {M(1)}));
 
   auto dc = result->at(2)->asDiscreteConditional();
-  DiscreteValues mode;
+  DiscreteValues dv;
-  mode[C(1)] = 0;
+  dv[M(1)] = 0;
-  EXPECT_DOUBLES_EQUAL(0.6225, (*dc)(mode), 1e-3);
+  EXPECT_DOUBLES_EQUAL(1, dc->operator()(dv), 1e-3);
 }
 
 /* ************************************************************************* */
 TEST(HybridGaussianFactorGraph, eliminateFullSequentialSimple) {
   HybridGaussianFactorGraph hfg;
 
-  DiscreteKey c1(C(1), 2);
+  DiscreteKey m1(M(1), 2);
 
   // Add prior on x0
   hfg.add(JacobianFactor(X(0), I_3x3, Z_3x1));
   // Add factor between x0 and x1
   hfg.add(JacobianFactor(X(0), I_3x3, X(1), -I_3x3, Z_3x1));
 
-  DecisionTree<Key, GaussianFactor::shared_ptr> dt(
+  std::vector<GaussianFactor::shared_ptr> factors = {
-      C(1), boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
+      boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
-      boost::make_shared<JacobianFactor>(X(1), I_3x3, Vector3::Ones()));
+      boost::make_shared<JacobianFactor>(X(1), I_3x3, Vector3::Ones())};
-  hfg.add(GaussianMixtureFactor({X(1)}, {c1}, dt));
+  hfg.add(GaussianMixtureFactor({X(1)}, {m1}, factors));
 
   // Discrete probability table for c1
-  hfg.add(HybridDiscreteFactor(DecisionTreeFactor(c1, {2, 8})));
+  hfg.add(DecisionTreeFactor(m1, {2, 8}));
   // Joint discrete probability table for c1, c2
-  hfg.add(HybridDiscreteFactor(
+  hfg.add(DecisionTreeFactor({{M(1), 2}, {M(2), 2}}, "1 2 3 4"));
-      DecisionTreeFactor({{C(1), 2}, {C(2), 2}}, "1 2 3 4")));
 
-  auto result = hfg.eliminateSequential(
+  HybridBayesNet::shared_ptr result = hfg.eliminateSequential(
-      Ordering::ColamdConstrainedLast(hfg, {C(1), C(2)}));
+      Ordering::ColamdConstrainedLast(hfg, {M(1), M(2)}));
 
+  // There are 4 variables (2 continuous + 2 discrete) in the bayes net.
   EXPECT_LONGS_EQUAL(4, result->size());
 }
 
@@ -169,31 +169,33 @@ TEST(HybridGaussianFactorGraph, eliminateFullSequentialSimple) {
 TEST(HybridGaussianFactorGraph, eliminateFullMultifrontalSimple) {
   HybridGaussianFactorGraph hfg;
 
-  DiscreteKey c1(C(1), 2);
+  DiscreteKey m1(M(1), 2);
 
   hfg.add(JacobianFactor(X(0), I_3x3, Z_3x1));
   hfg.add(JacobianFactor(X(0), I_3x3, X(1), -I_3x3, Z_3x1));
 
   hfg.add(GaussianMixtureFactor::FromFactors(
-      {X(1)}, {{C(1), 2}},
+      {X(1)}, {{M(1), 2}},
       {boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
        boost::make_shared<JacobianFactor>(X(1), I_3x3, Vector3::Ones())}));
 
-  hfg.add(DecisionTreeFactor(c1, {2, 8}));
+  hfg.add(DecisionTreeFactor(m1, {2, 8}));
-  hfg.add(DecisionTreeFactor({{C(1), 2}, {C(2), 2}}, "1 2 3 4"));
+  hfg.add(DecisionTreeFactor({{M(1), 2}, {M(2), 2}}, "1 2 3 4"));
 
-  auto result = hfg.eliminateMultifrontal(
+  HybridBayesTree::shared_ptr result = hfg.eliminateMultifrontal(
-      Ordering::ColamdConstrainedLast(hfg, {C(1), C(2)}));
+      Ordering::ColamdConstrainedLast(hfg, {M(1), M(2)}));
 
+  // The bayes tree should have 3 cliques
+  EXPECT_LONGS_EQUAL(3, result->size());
   // GTSAM_PRINT(*result);
-  // GTSAM_PRINT(*result->marginalFactor(C(2)));
+  // GTSAM_PRINT(*result->marginalFactor(M(2)));
 }
 
 /* ************************************************************************* */
 TEST(HybridGaussianFactorGraph, eliminateFullMultifrontalCLG) {
   HybridGaussianFactorGraph hfg;
 
-  DiscreteKey c(C(1), 2);
+  DiscreteKey m(M(1), 2);
 
   // Prior on x0
   hfg.add(JacobianFactor(X(0), I_3x3, Z_3x1));
@@ -202,16 +204,16 @@ TEST(HybridGaussianFactorGraph, eliminateFullMultifrontalCLG) {
 
   // Decision tree with different modes on x1
   DecisionTree<Key, GaussianFactor::shared_ptr> dt(
-      C(1), boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
+      M(1), boost::make_shared<JacobianFactor>(X(1), I_3x3, Z_3x1),
       boost::make_shared<JacobianFactor>(X(1), I_3x3, Vector3::Ones()));
 
   // Hybrid factor P(x1|c1)
-  hfg.add(GaussianMixtureFactor({X(1)}, {c}, dt));
+  hfg.add(GaussianMixtureFactor({X(1)}, {m}, dt));
   // Prior factor on c1
-  hfg.add(HybridDiscreteFactor(DecisionTreeFactor(c, {2, 8})));
+  hfg.add(HybridDiscreteFactor(DecisionTreeFactor(m, {2, 8})));
 
   // Get a constrained ordering keeping c1 last
-  auto ordering_full = Ordering::ColamdConstrainedLast(hfg, {C(1)});
+  auto ordering_full = Ordering::ColamdConstrainedLast(hfg, {M(1)});
 
   // Returns a Hybrid Bayes Tree with distribution P(x0|x1)P(x1|c1)P(c1)
   HybridBayesTree::shared_ptr hbt = hfg.eliminateMultifrontal(ordering_full);
@@ -232,48 +234,48 @@ TEST(HybridGaussianFactorGraph, eliminateFullMultifrontalTwoClique) {
 
   {
     hfg.add(GaussianMixtureFactor::FromFactors(
-        {X(0)}, {{C(0), 2}},
+        {X(0)}, {{M(0), 2}},
         {boost::make_shared<JacobianFactor>(X(0), I_3x3, Z_3x1),
          boost::make_shared<JacobianFactor>(X(0), I_3x3, Vector3::Ones())}));
 
     DecisionTree<Key, GaussianFactor::shared_ptr> dt1(
-        C(1), boost::make_shared<JacobianFactor>(X(2), I_3x3, Z_3x1),
+        M(1), boost::make_shared<JacobianFactor>(X(2), I_3x3, Z_3x1),
         boost::make_shared<JacobianFactor>(X(2), I_3x3, Vector3::Ones()));
 
-    hfg.add(GaussianMixtureFactor({X(2)}, {{C(1), 2}}, dt1));
+    hfg.add(GaussianMixtureFactor({X(2)}, {{M(1), 2}}, dt1));
   }
 
   hfg.add(HybridDiscreteFactor(
-      DecisionTreeFactor({{C(1), 2}, {C(2), 2}}, "1 2 3 4")));
+      DecisionTreeFactor({{M(1), 2}, {M(2), 2}}, "1 2 3 4")));
 
   hfg.add(JacobianFactor(X(3), I_3x3, X(4), -I_3x3, Z_3x1));
   hfg.add(JacobianFactor(X(4), I_3x3, X(5), -I_3x3, Z_3x1));
 
   {
     DecisionTree<Key, GaussianFactor::shared_ptr> dt(
-        C(3), boost::make_shared<JacobianFactor>(X(3), I_3x3, Z_3x1),
+        M(3), boost::make_shared<JacobianFactor>(X(3), I_3x3, Z_3x1),
         boost::make_shared<JacobianFactor>(X(3), I_3x3, Vector3::Ones()));
 
-    hfg.add(GaussianMixtureFactor({X(3)}, {{C(3), 2}}, dt));
+    hfg.add(GaussianMixtureFactor({X(3)}, {{M(3), 2}}, dt));
 
     DecisionTree<Key, GaussianFactor::shared_ptr> dt1(
-        C(2), boost::make_shared<JacobianFactor>(X(5), I_3x3, Z_3x1),
+        M(2), boost::make_shared<JacobianFactor>(X(5), I_3x3, Z_3x1),
         boost::make_shared<JacobianFactor>(X(5), I_3x3, Vector3::Ones()));
 
-    hfg.add(GaussianMixtureFactor({X(5)}, {{C(2), 2}}, dt1));
+    hfg.add(GaussianMixtureFactor({X(5)}, {{M(2), 2}}, dt1));
   }
 
   auto ordering_full =
-      Ordering::ColamdConstrainedLast(hfg, {C(0), C(1), C(2), C(3)});
+      Ordering::ColamdConstrainedLast(hfg, {M(0), M(1), M(2), M(3)});
 
   HybridBayesTree::shared_ptr hbt;
   HybridGaussianFactorGraph::shared_ptr remaining;
   std::tie(hbt, remaining) = hfg.eliminatePartialMultifrontal(ordering_full);
 
-  // GTSAM_PRINT(*hbt);
+  // 9 cliques in the bayes tree and 0 remaining variables to eliminate.
-  // GTSAM_PRINT(*remaining);
+  EXPECT_LONGS_EQUAL(9, hbt->size());
+  EXPECT_LONGS_EQUAL(0, remaining->size());
 
-  // hbt->marginalFactor(X(1))->print("HBT: ");
   /*
   (Fan) Explanation: the Junction tree will need to reeliminate to get to the
   marginal on X(1), which is not possible because it involves eliminating
@@ -307,13 +309,13 @@ TEST(HybridGaussianFactorGraph, Switching) {
   // X(3), X(7)
   // X(2), X(8)
   // X(1), X(4), X(6), X(9)
-  // C(5) will be the center, C(1-4), C(6-8)
+  // M(5) will be the center, M(1-4), M(6-8)
-  // C(3), C(7)
+  // M(3), M(7)
-  // C(1), C(4), C(2), C(6), C(8)
+  // M(1), M(4), M(2), M(6), M(8)
   // auto ordering_full =
   //     Ordering(KeyVector{X(1), X(4), X(2), X(6), X(9), X(8), X(3), X(7),
   //     X(5),
-  //                        C(1), C(4), C(2), C(6), C(8), C(3), C(7), C(5)});
+  //                        M(1), M(4), M(2), M(6), M(8), M(3), M(7), M(5)});
   KeyVector ordering;
 
   {
@@ -336,7 +338,7 @@ TEST(HybridGaussianFactorGraph, Switching) {
     std::iota(naturalC.begin(), naturalC.end(), 1);
     std::vector<Key> ordC;
     std::transform(naturalC.begin(), naturalC.end(), std::back_inserter(ordC),
-                   [](int x) { return C(x); });
+                   [](int x) { return M(x); });
     KeyVector ndC;
     std::vector<int> lvls;
 
@@ -353,9 +355,9 @@ TEST(HybridGaussianFactorGraph, Switching) {
   HybridGaussianFactorGraph::shared_ptr remaining;
   std::tie(hbt, remaining) = hfg->eliminatePartialMultifrontal(ordering_full);
 
-  // GTSAM_PRINT(*hbt);
+  // 12 cliques in the bayes tree and 0 remaining variables to eliminate.
-  // GTSAM_PRINT(*remaining);
+  EXPECT_LONGS_EQUAL(12, hbt->size());
-  // hbt->marginalFactor(C(11))->print("HBT: ");
+  EXPECT_LONGS_EQUAL(0, remaining->size());
 }
 
 /* ************************************************************************* */
@@ -368,13 +370,13 @@ TEST(HybridGaussianFactorGraph, SwitchingISAM) {
   // X(3), X(7)
   // X(2), X(8)
   // X(1), X(4), X(6), X(9)
-  // C(5) will be the center, C(1-4), C(6-8)
+  // M(5) will be the center, M(1-4), M(6-8)
-  // C(3), C(7)
+  // M(3), M(7)
-  // C(1), C(4), C(2), C(6), C(8)
+  // M(1), M(4), M(2), M(6), M(8)
   // auto ordering_full =
   //     Ordering(KeyVector{X(1), X(4), X(2), X(6), X(9), X(8), X(3), X(7),
   //     X(5),
-  //                        C(1), C(4), C(2), C(6), C(8), C(3), C(7), C(5)});
+  //                        M(1), M(4), M(2), M(6), M(8), M(3), M(7), M(5)});
   KeyVector ordering;
 
   {
@@ -397,7 +399,7 @@ TEST(HybridGaussianFactorGraph, SwitchingISAM) {
     std::iota(naturalC.begin(), naturalC.end(), 1);
     std::vector<Key> ordC;
     std::transform(naturalC.begin(), naturalC.end(), std::back_inserter(ordC),
-                   [](int x) { return C(x); });
+                   [](int x) { return M(x); });
     KeyVector ndC;
     std::vector<int> lvls;
 
@@ -407,9 +409,6 @@ TEST(HybridGaussianFactorGraph, SwitchingISAM) {
   }
   auto ordering_full = Ordering(ordering);
 
-  // GTSAM_PRINT(*hfg);
-  // GTSAM_PRINT(ordering_full);
-
   HybridBayesTree::shared_ptr hbt;
   HybridGaussianFactorGraph::shared_ptr remaining;
   std::tie(hbt, remaining) = hfg->eliminatePartialMultifrontal(ordering_full);
@@ -417,19 +416,18 @@ TEST(HybridGaussianFactorGraph, SwitchingISAM) {
   auto new_fg = makeSwitchingChain(12);
   auto isam = HybridGaussianISAM(*hbt);
 
-  {
+  // Run an ISAM update.
-    HybridGaussianFactorGraph factorGraph;
+  HybridGaussianFactorGraph factorGraph;
-    factorGraph.push_back(new_fg->at(new_fg->size() - 2));
+  factorGraph.push_back(new_fg->at(new_fg->size() - 2));
-    factorGraph.push_back(new_fg->at(new_fg->size() - 1));
+  factorGraph.push_back(new_fg->at(new_fg->size() - 1));
-    isam.update(factorGraph);
+  isam.update(factorGraph);
-    // std::cout << isam.dot();
+
-    // isam.marginalFactor(C(11))->print();
+  // ISAM should have 12 factors after the last update
-  }
+  EXPECT_LONGS_EQUAL(12, isam.size());
 }
 
 /* ************************************************************************* */
-// TODO(Varun) Actually test something!
+TEST(HybridGaussianFactorGraph, SwitchingTwoVar) {
-TEST_DISABLED(HybridGaussianFactorGraph, SwitchingTwoVar) {
   const int N = 7;
   auto hfg = makeSwitchingChain(N, X);
   hfg->push_back(*makeSwitchingChain(N, Y, D));
@@ -449,7 +447,7 @@ TEST_DISABLED(HybridGaussianFactorGraph, SwitchingTwoVar) {
   }
 
   for (size_t i = 1; i <= N - 1; i++) {
-    ordX.emplace_back(C(i));
+    ordX.emplace_back(M(i));
   }
   for (size_t i = 1; i <= N - 1; i++) {
     ordX.emplace_back(D(i));
@@ -461,8 +459,8 @@ TEST_DISABLED(HybridGaussianFactorGraph, SwitchingTwoVar) {
     dw.positionHints['c'] = 0;
     dw.positionHints['d'] = 3;
     dw.positionHints['y'] = 2;
-    std::cout << hfg->dot(DefaultKeyFormatter, dw);
+    // std::cout << hfg->dot(DefaultKeyFormatter, dw);
-    std::cout << "\n";
+    // std::cout << "\n";
   }
 
   {
@@ -471,10 +469,10 @@ TEST_DISABLED(HybridGaussianFactorGraph, SwitchingTwoVar) {
     // dw.positionHints['c'] = 0;
     // dw.positionHints['d'] = 3;
     dw.positionHints['x'] = 1;
-    std::cout << "\n";
+    // std::cout << "\n";
     // std::cout << hfg->eliminateSequential(Ordering(ordX))
     //                  ->dot(DefaultKeyFormatter, dw);
-    hfg->eliminateMultifrontal(Ordering(ordX))->dot(std::cout);
+    // hfg->eliminateMultifrontal(Ordering(ordX))->dot(std::cout);
   }
 
   Ordering ordering_partial;
@@ -482,22 +480,22 @@ TEST_DISABLED(HybridGaussianFactorGraph, SwitchingTwoVar) {
     ordering_partial.emplace_back(X(i));
     ordering_partial.emplace_back(Y(i));
   }
-  {
+  HybridBayesNet::shared_ptr hbn;
-    HybridBayesNet::shared_ptr hbn;
+  HybridGaussianFactorGraph::shared_ptr remaining;
-    HybridGaussianFactorGraph::shared_ptr remaining;
+  std::tie(hbn, remaining) =
-    std::tie(hbn, remaining) =
+      hfg->eliminatePartialSequential(ordering_partial);
-        hfg->eliminatePartialSequential(ordering_partial);
 
-    // remaining->print();
+  EXPECT_LONGS_EQUAL(14, hbn->size());
-    {
+  EXPECT_LONGS_EQUAL(11, remaining->size());
-      DotWriter dw;
+
-      dw.positionHints['x'] = 1;
+  {
-      dw.positionHints['c'] = 0;
+    DotWriter dw;
-      dw.positionHints['d'] = 3;
+    dw.positionHints['x'] = 1;
-      dw.positionHints['y'] = 2;
+    dw.positionHints['c'] = 0;
-      std::cout << remaining->dot(DefaultKeyFormatter, dw);
+    dw.positionHints['d'] = 3;
-      std::cout << "\n";
+    dw.positionHints['y'] = 2;
-    }
+    // std::cout << remaining->dot(DefaultKeyFormatter, dw);
+    // std::cout << "\n";
   }
 }
 

gtsam/hybrid/tests/testGaussianMixtureFactor.cpp

@@ -74,7 +74,7 @@ TEST(GaussianMixtureFactor, Sum) {
   // Check that number of keys is 3
   EXPECT_LONGS_EQUAL(3, mixtureFactorA.keys().size());
 
-  // Check that number of discrete keys is 1 // TODO(Frank): should not exist?
+  // Check that number of discrete keys is 1
   EXPECT_LONGS_EQUAL(1, mixtureFactorA.discreteKeys().size());
 
   // Create sum of two mixture factors: it will be a decision tree now on both
@@ -104,7 +104,7 @@ TEST(GaussianMixtureFactor, Printing) {
   GaussianMixtureFactor mixtureFactor({X(1), X(2)}, {m1}, factors);
 
   std::string expected =
-      R"(Hybrid x1 x2; 1 ]{
+      R"(Hybrid [x1 x2; 1]{
  Choice(1) 
  0 Leaf :
   A[x1] = [

gtsam/hybrid/tests/testHybridFactorGraph.cpp

@@ -0,0 +1,45 @@
+/* ----------------------------------------------------------------------------
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+ * See LICENSE for the license information
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file    testHybridFactorGraph.cpp
+ * @brief   Unit tests for HybridFactorGraph
+ * @author  Varun Agrawal
+ * @date    May 2021
+ */
+
+#include <CppUnitLite/Test.h>
+#include <CppUnitLite/TestHarness.h>
+#include <gtsam/base/TestableAssertions.h>
+#include <gtsam/base/utilities.h>
+#include <gtsam/hybrid/HybridFactorGraph.h>
+#include <gtsam/inference/Symbol.h>
+#include <gtsam/nonlinear/PriorFactor.h>
+
+using namespace std;
+using namespace gtsam;
+using noiseModel::Isotropic;
+using symbol_shorthand::L;
+using symbol_shorthand::M;
+using symbol_shorthand::X;
+
+/* ****************************************************************************
+ * Test that any linearizedFactorGraph gaussian factors are appended to the
+ * existing gaussian factor graph in the hybrid factor graph.
+ */
+TEST(HybridFactorGraph, Constructor) {
+  // Initialize the hybrid factor graph
+  HybridFactorGraph fg;
+}
+
+/* ************************************************************************* */
+int main() {
+  TestResult tr;
+  return TestRegistry::runAllTests(tr);
+}
+/* ************************************************************************* */

gtsam/hybrid/tests/testHybridIncremental.cpp

@@ -0,0 +1,649 @@
+/* ----------------------------------------------------------------------------
+
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+
+ * See LICENSE for the license information
+
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file    testHybridIncremental.cpp
+ * @brief   Unit tests for incremental inference
+ * @author  Fan Jiang, Varun Agrawal, Frank Dellaert
+ * @date    Jan 2021
+ */
+
+#include <gtsam/discrete/DiscreteBayesNet.h>
+#include <gtsam/discrete/DiscreteDistribution.h>
+#include <gtsam/geometry/Pose2.h>
+#include <gtsam/hybrid/HybridGaussianISAM.h>
+#include <gtsam/linear/GaussianBayesNet.h>
+#include <gtsam/linear/GaussianFactorGraph.h>
+#include <gtsam/nonlinear/PriorFactor.h>
+#include <gtsam/sam/BearingRangeFactor.h>
+
+#include <numeric>
+
+#include "Switching.h"
+
+// Include for test suite
+#include <CppUnitLite/TestHarness.h>
+
+using namespace std;
+using namespace gtsam;
+using noiseModel::Isotropic;
+using symbol_shorthand::L;
+using symbol_shorthand::M;
+using symbol_shorthand::W;
+using symbol_shorthand::X;
+using symbol_shorthand::Y;
+using symbol_shorthand::Z;
+
+/* ****************************************************************************/
+// Test if we can perform elimination incrementally.
+TEST(HybridGaussianElimination, IncrementalElimination) {
+  Switching switching(3);
+  HybridGaussianISAM isam;
+  HybridGaussianFactorGraph graph1;
+
+  // Create initial factor graph
+  //    *        *      *
+  //    |        |      |
+  // *- X1  -*-  X2 -*- X3
+  //     \*-M1-*/
+  graph1.push_back(switching.linearizedFactorGraph.at(0));  // P(X1)
+  graph1.push_back(switching.linearizedFactorGraph.at(1));  // P(X1, X2 | M1)
+  graph1.push_back(switching.linearizedFactorGraph.at(2));  // P(X2, X3 | M2)
+  graph1.push_back(switching.linearizedFactorGraph.at(5));  // P(M1)
+
+  // Create ordering.
+  Ordering ordering;
+  ordering += X(1);
+  ordering += X(2);
+
+  // Run update step
+  isam.update(graph1);
+
+  // Check that after update we have 3 hybrid Bayes net nodes:
+  // P(X1 | X2, M1) and P(X2, X3 | M1, M2), P(M1, M2)
+  EXPECT_LONGS_EQUAL(3, isam.size());
+  EXPECT(isam[X(1)]->conditional()->frontals() == KeyVector{X(1)});
+  EXPECT(isam[X(1)]->conditional()->parents() == KeyVector({X(2), M(1)}));
+  EXPECT(isam[X(2)]->conditional()->frontals() == KeyVector({X(2), X(3)}));
+  EXPECT(isam[X(2)]->conditional()->parents() == KeyVector({M(1), M(2)}));
+
+  /********************************************************/
+  // New factor graph for incremental update.
+  HybridGaussianFactorGraph graph2;
+
+  graph1.push_back(switching.linearizedFactorGraph.at(3));  // P(X2)
+  graph2.push_back(switching.linearizedFactorGraph.at(4));  // P(X3)
+  graph2.push_back(switching.linearizedFactorGraph.at(6));  // P(M1, M2)
+
+  // Create ordering.
+  Ordering ordering2;
+  ordering2 += X(2);
+  ordering2 += X(3);
+
+  isam.update(graph2);
+
+  // Check that after the second update we have
+  // 1 additional hybrid Bayes net node:
+  // P(X2, X3 | M1, M2)
+  EXPECT_LONGS_EQUAL(3, isam.size());
+  EXPECT(isam[X(3)]->conditional()->frontals() == KeyVector({X(2), X(3)}));
+  EXPECT(isam[X(3)]->conditional()->parents() == KeyVector({M(1), M(2)}));
+}
+
+/* ****************************************************************************/
+// Test if we can incrementally do the inference
+TEST(HybridGaussianElimination, IncrementalInference) {
+  Switching switching(3);
+  HybridGaussianISAM isam;
+  HybridGaussianFactorGraph graph1;
+
+  // Create initial factor graph
+  //    *        *        *
+  //    |        |        |
+  // *- X1  -*-  X2  -*-  X3
+  //         |        |
+  //      *-M1 - * - M2
+  graph1.push_back(switching.linearizedFactorGraph.at(0));  // P(X1)
+  graph1.push_back(switching.linearizedFactorGraph.at(1));  // P(X1, X2 | M1)
+  graph1.push_back(switching.linearizedFactorGraph.at(3));  // P(X2)
+  graph1.push_back(switching.linearizedFactorGraph.at(5));  // P(M1)
+
+  //TODO(Varun) we cannot enforce ordering
+  // // Create ordering.
+  // Ordering ordering1;
+  // ordering1 += X(1);
+  // ordering1 += X(2);
+
+  // Run update step
+  isam.update(graph1);
+
+  /********************************************************/
+  // New factor graph for incremental update.
+  HybridGaussianFactorGraph graph2;
+
+  graph2.push_back(switching.linearizedFactorGraph.at(2));  // P(X2, X3 | M2)
+  graph2.push_back(switching.linearizedFactorGraph.at(4));  // P(X3)
+  graph2.push_back(switching.linearizedFactorGraph.at(6));  // P(M1, M2)
+
+  //TODO(Varun) we cannot enforce ordering
+  // // Create ordering.
+  // Ordering ordering2;
+  // ordering2 += X(2);
+  // ordering2 += X(3);
+
+  isam.update(graph2);
+  GTSAM_PRINT(isam);
+
+  /********************************************************/
+  // Run batch elimination so we can compare results.
+  Ordering ordering;
+  ordering += X(1);
+  ordering += X(2);
+  ordering += X(3);
+
+  // Now we calculate the actual factors using full elimination
+  HybridBayesNet::shared_ptr expectedHybridBayesNet;
+  HybridGaussianFactorGraph::shared_ptr expectedRemainingGraph;
+  std::tie(expectedHybridBayesNet, expectedRemainingGraph) =
+      switching.linearizedFactorGraph.eliminatePartialSequential(ordering);
+
+  // The densities on X(1) should be the same
+  auto x1_conditional =
+      dynamic_pointer_cast<GaussianMixture>(isam[X(1)]->conditional()->inner());
+  EXPECT(
+      assert_equal(*x1_conditional, *(expectedHybridBayesNet->atGaussian(0))));
+
+  // The densities on X(2) should be the same
+  auto x2_conditional =
+      dynamic_pointer_cast<GaussianMixture>(isam[X(2)]->conditional()->inner());
+  EXPECT(
+      assert_equal(*x2_conditional, *(expectedHybridBayesNet->atGaussian(1))));
+
+  // // The densities on X(3) should be the same
+  // auto x3_conditional =
+  //     dynamic_pointer_cast<GaussianMixture>(isam[X(3)]->conditional()->inner());
+  // EXPECT(
+  //     assert_equal(*x3_conditional, *(expectedHybridBayesNet->atGaussian(2))));
+
+  GTSAM_PRINT(*expectedHybridBayesNet);
+
+  // we only do the manual continuous elimination for 0,0
+  // the other discrete probabilities on M(2) are calculated the same way
+  auto m00_prob = [&]() {
+    GaussianFactorGraph gf;
+    // gf.add(switching.linearizedFactorGraph.gaussianGraph().at(3));
+
+    DiscreteValues m00;
+    m00[M(1)] = 0, m00[M(2)] = 0;
+    // auto dcMixture =
+    //     dynamic_pointer_cast<DCGaussianMixtureFactor>(graph2.dcGraph().at(0));
+    // gf.add(dcMixture->factors()(m00));
+    // auto x2_mixed =
+    //     boost::dynamic_pointer_cast<GaussianMixture>(hybridBayesNet.at(1));
+    // gf.add(x2_mixed->factors()(m00));
+    auto result_gf = gf.eliminateSequential();
+    return gf.probPrime(result_gf->optimize());
+  }();
+
+  /// Test if the probability values are as expected with regression tests.
+//   DiscreteValues assignment;
+//   EXPECT(assert_equal(m00_prob, 0.60656, 1e-5));
+//   assignment[M(1)] = 0;
+//   assignment[M(2)] = 0;
+//   EXPECT(assert_equal(m00_prob, (*discreteFactor)(assignment), 1e-5));
+//   assignment[M(1)] = 1;
+//   assignment[M(2)] = 0;
+//   EXPECT(assert_equal(0.612477, (*discreteFactor)(assignment), 1e-5));
+//   assignment[M(1)] = 0;
+//   assignment[M(2)] = 1;
+//   EXPECT(assert_equal(0.999952, (*discreteFactor)(assignment), 1e-5));
+//   assignment[M(1)] = 1;
+//   assignment[M(2)] = 1;
+//   EXPECT(assert_equal(1.0, (*discreteFactor)(assignment), 1e-5));
+
+//   DiscreteFactorGraph dfg;
+//   dfg.add(*discreteFactor);
+//   dfg.add(discreteFactor_m1);
+//   dfg.add_factors(switching.linearizedFactorGraph.discreteGraph());
+
+//   // Check if the chordal graph generated from incremental elimination
+//   matches
+//   // that of batch elimination.
+//   auto chordal = dfg.eliminateSequential();
+//   auto expectedChordal =
+//       expectedRemainingGraph->discreteGraph().eliminateSequential();
+
+//   EXPECT(assert_equal(*expectedChordal, *chordal, 1e-6));
+}
+
+/* ****************************************************************************/
+// // Test if we can approximately do the inference
+// TEST(DCGaussianElimination, Approx_inference) {
+//   Switching switching(4);
+//   IncrementalHybrid incrementalHybrid;
+//   HybridGaussianFactorGraph graph1;
+
+//   // Add the 3 DC factors, x1-x2, x2-x3, x3-x4
+//   for (size_t i = 0; i < 3; i++) {
+//     graph1.push_back(switching.linearizedFactorGraph.dcGraph().at(i));
+//   }
+
+//   // Add the Gaussian factors, 1 prior on X(1), 4 measurements
+//   for (size_t i = 0; i <= 4; i++) {
+// graph1.push_back(switching.linearizedFactorGraph.gaussianGraph().at(i));
+//   }
+
+//   // Create ordering.
+//   Ordering ordering;
+//   for (size_t j = 1; j <= 4; j++) {
+//     ordering += X(j);
+//   }
+
+//   // Now we calculate the actual factors using full elimination
+//   HybridBayesNet::shared_ptr unprunedHybridBayesNet;
+//   HybridGaussianFactorGraph::shared_ptr unprunedRemainingGraph;
+//   std::tie(unprunedHybridBayesNet, unprunedRemainingGraph) =
+// switching.linearizedFactorGraph.eliminatePartialSequential(ordering);
+
+//   size_t maxComponents = 5;
+//   incrementalHybrid.update(graph1, ordering, maxComponents);
+
+//   /*
+//   unpruned factor is:
+//       Choice(m3)
+//       0 Choice(m2)
+//       0 0 Choice(m1)
+//       0 0 0 Leaf 0.2248 -
+//       0 0 1 Leaf 0.3715 -
+//       0 1 Choice(m1)
+//       0 1 0 Leaf 0.3742 *
+//       0 1 1 Leaf 0.6125 *
+//       1 Choice(m2)
+//       1 0 Choice(m1)
+//       1 0 0 Leaf 0.3706 -
+//       1 0 1 Leaf 0.6124 *
+//       1 1 Choice(m1)
+//       1 1 0 Leaf 0.611 *
+//       1 1 1 Leaf    1 *
+
+//   pruned factors is:
+//       Choice(m3)
+//       0 Choice(m2)
+//       0 0 Leaf    0
+//       0 1 Choice(m1)
+//       0 1 0 Leaf 0.3742
+//       0 1 1 Leaf 0.6125
+//       1 Choice(m2)
+//       1 0 Choice(m1)
+//       1 0 0 Leaf    0
+//       1 0 1 Leaf 0.6124
+//       1 1 Choice(m1)
+//       1 1 0 Leaf 0.611
+//       1 1 1 Leaf    1
+//    */
+
+//   // Test that the remaining factor graph has one
+//   // DecisionTreeFactor on {M3, M2, M1}.
+//   auto remainingFactorGraph = incrementalHybrid.remainingFactorGraph();
+//   EXPECT_LONGS_EQUAL(1, remainingFactorGraph.size());
+
+//   auto discreteFactor_m1 = *dynamic_pointer_cast<DecisionTreeFactor>(
+//       incrementalHybrid.remainingDiscreteGraph().at(0));
+//   EXPECT(discreteFactor_m1.keys() == KeyVector({M(3), M(2), M(1)}));
+
+//   // Get the number of elements which are equal to 0.
+//   auto count = [](const double &value, int count) {
+//     return value > 0 ? count + 1 : count;
+//   };
+//   // Check that the number of leaves after pruning is 5.
+//   EXPECT_LONGS_EQUAL(5, discreteFactor_m1.fold(count, 0));
+
+//   /* Expected hybrid Bayes net
+//    * factor 0:  [x1 | x2 m1 ], 2 components
+//    * factor 1:  [x2 | x3 m2 m1 ], 4 components
+//    * factor 2:  [x3 | x4 m3 m2 m1 ], 8 components
+//    * factor 3:  [x4 | m3 m2 m1 ], 8 components
+//    */
+//   auto hybridBayesNet = incrementalHybrid.hybridBayesNet();
+
+//   // Check if we have a bayes net with 4 hybrid nodes,
+//   // each with 2, 4, 5 (pruned), and 5 (pruned) leaves respetively.
+//   EXPECT_LONGS_EQUAL(4, hybridBayesNet.size());
+//   EXPECT_LONGS_EQUAL(2, hybridBayesNet.atGaussian(0)->nrComponents());
+//   EXPECT_LONGS_EQUAL(4, hybridBayesNet.atGaussian(1)->nrComponents());
+//   EXPECT_LONGS_EQUAL(5, hybridBayesNet.atGaussian(2)->nrComponents());
+//   EXPECT_LONGS_EQUAL(5, hybridBayesNet.atGaussian(3)->nrComponents());
+
+//   // Check that the hybrid nodes of the bayes net match those of the bayes
+//   net
+//   // before pruning, at the same positions.
+//   auto &lastDensity = *(hybridBayesNet.atGaussian(3));
+//   auto &unprunedLastDensity = *(unprunedHybridBayesNet->atGaussian(3));
+//   std::vector<std::pair<DiscreteValues, double>> assignments =
+//       discreteFactor_m1.enumerate();
+//   // Loop over all assignments and check the pruned components
+//   for (auto &&av : assignments) {
+//     const DiscreteValues &assignment = av.first;
+//     const double value = av.second;
+
+//     if (value == 0.0) {
+//       EXPECT(lastDensity(assignment) == nullptr);
+//     } else {
+//       CHECK(lastDensity(assignment));
+//       EXPECT(assert_equal(*unprunedLastDensity(assignment),
+//                           *lastDensity(assignment)));
+//     }
+//   }
+// }
+
+/* ****************************************************************************/
+// // Test approximate inference with an additional pruning step.
+// TEST(DCGaussianElimination, Incremental_approximate) {
+//   Switching switching(5);
+//   IncrementalHybrid incrementalHybrid;
+//   HybridGaussianFactorGraph graph1;
+
+//   // Add the 3 DC factors, x1-x2, x2-x3, x3-x4
+//   for (size_t i = 0; i < 3; i++) {
+//     graph1.push_back(switching.linearizedFactorGraph.dcGraph().at(i));
+//   }
+
+//   // Add the Gaussian factors, 1 prior on X(1), 4 measurements
+//   for (size_t i = 0; i <= 4; i++) {
+// graph1.push_back(switching.linearizedFactorGraph.gaussianGraph().at(i));
+//   }
+
+//   // Create ordering.
+//   Ordering ordering;
+//   for (size_t j = 1; j <= 4; j++) {
+//     ordering += X(j);
+//   }
+
+//   // Run update with pruning
+//   size_t maxComponents = 5;
+//   incrementalHybrid.update(graph1, ordering, maxComponents);
+
+//   // Check if we have a bayes net with 4 hybrid nodes,
+//   // each with 2, 4, 8, and 5 (pruned) leaves respetively.
+//   auto actualBayesNet1 = incrementalHybrid.hybridBayesNet();
+//   CHECK_EQUAL(4, actualBayesNet1.size());
+//   EXPECT_LONGS_EQUAL(2, actualBayesNet1.atGaussian(0)->nrComponents());
+//   EXPECT_LONGS_EQUAL(4, actualBayesNet1.atGaussian(1)->nrComponents());
+//   EXPECT_LONGS_EQUAL(8, actualBayesNet1.atGaussian(2)->nrComponents());
+//   EXPECT_LONGS_EQUAL(5, actualBayesNet1.atGaussian(3)->nrComponents());
+
+//   /***** Run Round 2 *****/
+//   HybridGaussianFactorGraph graph2;
+//   graph2.push_back(switching.linearizedFactorGraph.dcGraph().at(3));
+//   graph2.push_back(switching.linearizedFactorGraph.gaussianGraph().at(5));
+
+//   Ordering ordering2;
+//   ordering2 += X(4);
+//   ordering2 += X(5);
+
+//   // Run update with pruning a second time.
+//   incrementalHybrid.update(graph2, ordering2, maxComponents);
+
+//   // Check if we have a bayes net with 2 hybrid nodes,
+//   // each with 10 (pruned), and 5 (pruned) leaves respetively.
+//   auto actualBayesNet = incrementalHybrid.hybridBayesNet();
+//   CHECK_EQUAL(2, actualBayesNet.size());
+//   EXPECT_LONGS_EQUAL(10, actualBayesNet.atGaussian(0)->nrComponents());
+//   EXPECT_LONGS_EQUAL(5, actualBayesNet.atGaussian(1)->nrComponents());
+// }
+
+/* ************************************************************************/
+// // Test for figuring out the optimal ordering to ensure we get
+// // a discrete graph after elimination.
+// TEST(IncrementalHybrid, NonTrivial) {
+//   // This is a GTSAM-only test for running inference on a single legged
+//   robot.
+//   // The leg links are represented by the chain X-Y-Z-W, where X is the base
+//   and
+//   // W is the foot.
+//   // We use BetweenFactor<Pose2> as constraints between each of the poses.
+
+//   /*************** Run Round 1 ***************/
+//   NonlinearHybridFactorGraph fg;
+
+//   // Add a prior on pose x1 at the origin.
+//   // A prior factor consists of a mean  and
+//   // a noise model (covariance matrix)
+//   Pose2 prior(0.0, 0.0, 0.0);  // prior mean is at origin
+//   auto priorNoise = noiseModel::Diagonal::Sigmas(
+//       Vector3(0.3, 0.3, 0.1));  // 30cm std on x,y, 0.1 rad on theta
+//   fg.emplace_nonlinear<PriorFactor<Pose2>>(X(0), prior, priorNoise);
+
+//   // create a noise model for the landmark measurements
+//   auto poseNoise = noiseModel::Isotropic::Sigma(3, 0.1);
+
+//   // We model a robot's single leg as X - Y - Z - W
+//   // where X is the base link and W is the foot link.
+
+//   // Add connecting poses similar to PoseFactors in GTD
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(0), Y(0), Pose2(0, 1.0, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Y(0), Z(0), Pose2(0, 1.0, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Z(0), W(0), Pose2(0, 1.0, 0),
+//                                              poseNoise);
+
+//   // Create initial estimate
+//   Values initial;
+//   initial.insert(X(0), Pose2(0.0, 0.0, 0.0));
+//   initial.insert(Y(0), Pose2(0.0, 1.0, 0.0));
+//   initial.insert(Z(0), Pose2(0.0, 2.0, 0.0));
+//   initial.insert(W(0), Pose2(0.0, 3.0, 0.0));
+
+//   HybridGaussianFactorGraph gfg = fg.linearize(initial);
+//   fg = NonlinearHybridFactorGraph();
+
+//   IncrementalHybrid inc;
+
+//   // Regular ordering going up the chain.
+//   Ordering ordering;
+//   ordering += W(0);
+//   ordering += Z(0);
+//   ordering += Y(0);
+//   ordering += X(0);
+
+//   // Update without pruning
+//   // The result is a HybridBayesNet with no discrete variables
+//   // (equivalent to a GaussianBayesNet).
+//   // Factorization is:
+//   // `P(X | measurements) = P(W0|Z0) P(Z0|Y0) P(Y0|X0) P(X0)`
+//   inc.update(gfg, ordering);
+
+//   /*************** Run Round 2 ***************/
+//   using PlanarMotionModel = BetweenFactor<Pose2>;
+
+//   // Add odometry factor with discrete modes.
+//   Pose2 odometry(1.0, 0.0, 0.0);
+//   KeyVector contKeys = {W(0), W(1)};
+//   auto noise_model = noiseModel::Isotropic::Sigma(3, 1.0);
+//   auto still = boost::make_shared<PlanarMotionModel>(W(0), W(1), Pose2(0, 0,
+//   0),
+//                                                      noise_model),
+//        moving = boost::make_shared<PlanarMotionModel>(W(0), W(1), odometry,
+//                                                       noise_model);
+//   std::vector<PlanarMotionModel::shared_ptr> components = {moving, still};
+//   auto dcFactor = boost::make_shared<DCMixtureFactor<PlanarMotionModel>>(
+//       contKeys, DiscreteKeys{gtsam::DiscreteKey(M(1), 2)}, components);
+//   fg.push_back(dcFactor);
+
+//   // Add equivalent of ImuFactor
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(0), X(1), Pose2(1.0, 0.0, 0),
+//                                              poseNoise);
+//   // PoseFactors-like at k=1
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(1), Y(1), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Y(1), Z(1), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Z(1), W(1), Pose2(-1, 1, 0),
+//                                              poseNoise);
+
+//   initial.insert(X(1), Pose2(1.0, 0.0, 0.0));
+//   initial.insert(Y(1), Pose2(1.0, 1.0, 0.0));
+//   initial.insert(Z(1), Pose2(1.0, 2.0, 0.0));
+//   // The leg link did not move so we set the expected pose accordingly.
+//   initial.insert(W(1), Pose2(0.0, 3.0, 0.0));
+
+//   // Ordering for k=1.
+//   // This ordering follows the intuition that we eliminate the previous
+//   // timestep, and then the current timestep.
+//   ordering = Ordering();
+//   ordering += W(0);
+//   ordering += Z(0);
+//   ordering += Y(0);
+//   ordering += X(0);
+//   ordering += W(1);
+//   ordering += Z(1);
+//   ordering += Y(1);
+//   ordering += X(1);
+
+//   gfg = fg.linearize(initial);
+//   fg = NonlinearHybridFactorGraph();
+
+//   // Update without pruning
+//   // The result is a HybridBayesNet with 1 discrete variable M(1).
+//   // P(X | measurements) = P(W0|Z0, W1, M1) P(Z0|Y0, W1, M1) P(Y0|X0, W1, M1)
+//   //                       P(X0 | X1, W1, M1) P(W1|Z1, X1, M1) P(Z1|Y1, X1,
+//   M1)
+//   //                       P(Y1 | X1, M1)P(X1 | M1)P(M1)
+//   // The MHS tree is a 2 level tree for time indices (0, 1) with 2 leaves.
+//   inc.update(gfg, ordering);
+
+//   /*************** Run Round 3 ***************/
+//   // Add odometry factor with discrete modes.
+//   contKeys = {W(1), W(2)};
+//   still = boost::make_shared<PlanarMotionModel>(W(1), W(2), Pose2(0, 0, 0),
+//                                                 noise_model);
+//   moving =
+//       boost::make_shared<PlanarMotionModel>(W(1), W(2), odometry,
+//       noise_model);
+//   components = {moving, still};
+//   dcFactor = boost::make_shared<DCMixtureFactor<PlanarMotionModel>>(
+//       contKeys, DiscreteKeys{gtsam::DiscreteKey(M(2), 2)}, components);
+//   fg.push_back(dcFactor);
+
+//   // Add equivalent of ImuFactor
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(1), X(2), Pose2(1.0, 0.0, 0),
+//                                              poseNoise);
+//   // PoseFactors-like at k=1
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(2), Y(2), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Y(2), Z(2), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Z(2), W(2), Pose2(-2, 1, 0),
+//                                              poseNoise);
+
+//   initial.insert(X(2), Pose2(2.0, 0.0, 0.0));
+//   initial.insert(Y(2), Pose2(2.0, 1.0, 0.0));
+//   initial.insert(Z(2), Pose2(2.0, 2.0, 0.0));
+//   initial.insert(W(2), Pose2(0.0, 3.0, 0.0));
+
+//   // Ordering at k=2. Same intuition as before.
+//   ordering = Ordering();
+//   ordering += W(1);
+//   ordering += Z(1);
+//   ordering += Y(1);
+//   ordering += X(1);
+//   ordering += W(2);
+//   ordering += Z(2);
+//   ordering += Y(2);
+//   ordering += X(2);
+
+//   gfg = fg.linearize(initial);
+//   fg = NonlinearHybridFactorGraph();
+
+//   // Now we prune!
+//   // P(X | measurements) = P(W0|Z0, W1, M1) P(Z0|Y0, W1, M1) P(Y0|X0, W1, M1)
+//   P(X0 | X1, W1, M1)
+//   //                       P(W1|W2, Z1, X1, M1, M2) P(Z1| W2, Y1, X1, M1, M2)
+//   P(Y1 | W2, X1, M1, M2)
+//   //                       P(X1 | W2, X2, M1, M2) P(W2|Z2, X2, M1, M2)
+//   P(Z2|Y2, X2, M1, M2)
+//   //                       P(Y2 | X2, M1, M2)P(X2 | M1, M2) P(M1, M2)
+//   // The MHS at this point should be a 3 level tree on (0, 1, 2).
+//   // 0 has 2 choices, 1 has 4 choices and 2 has 4 choices.
+//   inc.update(gfg, ordering, 2);
+
+//   /*************** Run Round 4 ***************/
+//   // Add odometry factor with discrete modes.
+//   contKeys = {W(2), W(3)};
+//   still = boost::make_shared<PlanarMotionModel>(W(2), W(3), Pose2(0, 0, 0),
+//                                                 noise_model);
+//   moving =
+//       boost::make_shared<PlanarMotionModel>(W(2), W(3), odometry,
+//       noise_model);
+//   components = {moving, still};
+//   dcFactor = boost::make_shared<DCMixtureFactor<PlanarMotionModel>>(
+//       contKeys, DiscreteKeys{gtsam::DiscreteKey(M(3), 2)}, components);
+//   fg.push_back(dcFactor);
+
+//   // Add equivalent of ImuFactor
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(2), X(3), Pose2(1.0, 0.0, 0),
+//                                              poseNoise);
+//   // PoseFactors-like at k=3
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(X(3), Y(3), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Y(3), Z(3), Pose2(0, 1, 0),
+//                                              poseNoise);
+//   fg.emplace_nonlinear<BetweenFactor<Pose2>>(Z(3), W(3), Pose2(-3, 1, 0),
+//                                              poseNoise);
+
+//   initial.insert(X(3), Pose2(3.0, 0.0, 0.0));
+//   initial.insert(Y(3), Pose2(3.0, 1.0, 0.0));
+//   initial.insert(Z(3), Pose2(3.0, 2.0, 0.0));
+//   initial.insert(W(3), Pose2(0.0, 3.0, 0.0));
+
+//   // Ordering at k=3. Same intuition as before.
+//   ordering = Ordering();
+//   ordering += W(2);
+//   ordering += Z(2);
+//   ordering += Y(2);
+//   ordering += X(2);
+//   ordering += W(3);
+//   ordering += Z(3);
+//   ordering += Y(3);
+//   ordering += X(3);
+
+//   gfg = fg.linearize(initial);
+//   fg = NonlinearHybridFactorGraph();
+
+//   // Keep pruning!
+//   inc.update(gfg, ordering, 3);
+
+//   // The final discrete graph should not be empty since we have eliminated
+//   // all continuous variables.
+//   EXPECT(!inc.remainingDiscreteGraph().empty());
+
+//   // Test if the optimal discrete mode assignment is (1, 1, 1).
+//   DiscreteValues optimal_assignment =
+//   inc.remainingDiscreteGraph().optimize(); DiscreteValues
+//   expected_assignment; expected_assignment[M(1)] = 1;
+//   expected_assignment[M(2)] = 1;
+//   expected_assignment[M(3)] = 1;
+//   EXPECT(assert_equal(expected_assignment, optimal_assignment));
+
+//   // Test if pruning worked correctly by checking that we only have 3 leaves
+//   in
+//   // the last node.
+//   auto lastConditional = boost::dynamic_pointer_cast<GaussianMixture>(
+//       inc.hybridBayesNet().at(inc.hybridBayesNet().size() - 1));
+//   EXPECT_LONGS_EQUAL(3, lastConditional->nrComponents());
+// }
+
+/* ************************************************************************* */
+int main() {
+  TestResult tr;
+  return TestRegistry::runAllTests(tr);
+}

gtsam/hybrid/tests/testHybridNonlinearFactorGraph.cpp

@@ -0,0 +1,761 @@
+/* ----------------------------------------------------------------------------
+ * GTSAM Copyright 2010, Georgia Tech Research Corporation,
+ * Atlanta, Georgia 30332-0415
+ * All Rights Reserved
+ * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
+ * See LICENSE for the license information
+ * -------------------------------------------------------------------------- */
+
+/**
+ * @file    testHybridNonlinearFactorGraph.cpp
+ * @brief   Unit tests for HybridNonlinearFactorGraph
+ * @author  Varun Agrawal
+ * @author  Fan Jiang
+ * @author  Frank Dellaert
+ * @date    December 2021
+ */
+
+#include <gtsam/base/TestableAssertions.h>
+#include <gtsam/base/utilities.h>
+#include <gtsam/discrete/DiscreteBayesNet.h>
+#include <gtsam/discrete/DiscreteDistribution.h>
+#include <gtsam/discrete/DiscreteFactorGraph.h>
+#include <gtsam/geometry/Pose2.h>
+#include <gtsam/hybrid/HybridEliminationTree.h>
+#include <gtsam/hybrid/HybridFactor.h>
+#include <gtsam/hybrid/HybridNonlinearFactorGraph.h>
+#include <gtsam/hybrid/MixtureFactor.h>
+#include <gtsam/linear/GaussianBayesNet.h>
+#include <gtsam/linear/GaussianFactorGraph.h>
+#include <gtsam/nonlinear/NonlinearFactorGraph.h>
+#include <gtsam/nonlinear/PriorFactor.h>
+#include <gtsam/sam/BearingRangeFactor.h>
+#include <gtsam/slam/BetweenFactor.h>
+
+#include <numeric>
+
+#include "Switching.h"
+
+// Include for test suite
+#include <CppUnitLite/TestHarness.h>
+
+using namespace std;
+using namespace gtsam;
+using noiseModel::Isotropic;
+using symbol_shorthand::L;
+using symbol_shorthand::M;
+using symbol_shorthand::X;
+
+/* ****************************************************************************
+ * Test that any linearizedFactorGraph gaussian factors are appended to the
+ * existing gaussian factor graph in the hybrid factor graph.
+ */
+TEST(HybridFactorGraph, GaussianFactorGraph) {
+  HybridNonlinearFactorGraph fg;
+
+  // Add a simple prior factor to the nonlinear factor graph
+  fg.emplace_nonlinear<PriorFactor<double>>(X(0), 0, Isotropic::Sigma(1, 0.1));
+
+  // Linearization point
+  Values linearizationPoint;
+  linearizationPoint.insert<double>(X(0), 0);
+
+  HybridGaussianFactorGraph ghfg = fg.linearize(linearizationPoint);
+
+  // Add a factor to the GaussianFactorGraph
+  ghfg.add(JacobianFactor(X(0), I_1x1, Vector1(5)));
+
+  EXPECT_LONGS_EQUAL(2, ghfg.size());
+}
+
+/***************************************************************************
+ * Test equality for Hybrid Nonlinear Factor Graph.
+ */
+TEST(HybridNonlinearFactorGraph, Equals) {
+  HybridNonlinearFactorGraph graph1;
+  HybridNonlinearFactorGraph graph2;
+
+  // Test empty factor graphs
+  EXPECT(assert_equal(graph1, graph2));
+
+  auto f0 = boost::make_shared<PriorFactor<Pose2>>(
+      1, Pose2(), noiseModel::Isotropic::Sigma(3, 0.001));
+  graph1.push_back(f0);
+  graph2.push_back(f0);
+
+  auto f1 = boost::make_shared<BetweenFactor<Pose2>>(
+      1, 2, Pose2(), noiseModel::Isotropic::Sigma(3, 0.1));
+  graph1.push_back(f1);
+  graph2.push_back(f1);
+
+  // Test non-empty graphs
+  EXPECT(assert_equal(graph1, graph2));
+}
+
+/***************************************************************************
+ * Test that the resize method works correctly for a HybridNonlinearFactorGraph.
+ */
+TEST(HybridNonlinearFactorGraph, Resize) {
+  HybridNonlinearFactorGraph fg;
+  auto nonlinearFactor = boost::make_shared<BetweenFactor<double>>();
+  fg.push_back(nonlinearFactor);
+
+  auto discreteFactor = boost::make_shared<DecisionTreeFactor>();
+  fg.push_back(discreteFactor);
+
+  auto dcFactor = boost::make_shared<MixtureFactor>();
+  fg.push_back(dcFactor);
+
+  EXPECT_LONGS_EQUAL(fg.size(), 3);
+
+  fg.resize(0);
+  EXPECT_LONGS_EQUAL(fg.size(), 0);
+}
+
+/***************************************************************************
+ * Test that the resize method works correctly for a
+ * HybridGaussianFactorGraph.
+ */
+TEST(HybridGaussianFactorGraph, Resize) {
+  HybridNonlinearFactorGraph nhfg;
+  auto nonlinearFactor = boost::make_shared<BetweenFactor<double>>(
+      X(0), X(1), 0.0, Isotropic::Sigma(1, 0.1));
+  nhfg.push_back(nonlinearFactor);
+  auto discreteFactor = boost::make_shared<DecisionTreeFactor>();
+  nhfg.push_back(discreteFactor);
+
+  KeyVector contKeys = {X(0), X(1)};
+  auto noise_model = noiseModel::Isotropic::Sigma(1, 1.0);
+  auto still = boost::make_shared<MotionModel>(X(0), X(1), 0.0, noise_model),
+       moving = boost::make_shared<MotionModel>(X(0), X(1), 1.0, noise_model);
+
+  std::vector<MotionModel::shared_ptr> components = {still, moving};
+  auto dcFactor = boost::make_shared<MixtureFactor>(
+      contKeys, DiscreteKeys{gtsam::DiscreteKey(M(1), 2)}, components);
+  nhfg.push_back(dcFactor);
+
+  Values linearizationPoint;
+  linearizationPoint.insert<double>(X(0), 0);
+  linearizationPoint.insert<double>(X(1), 1);
+
+  // Generate `HybridGaussianFactorGraph` by linearizing
+  HybridGaussianFactorGraph gfg = nhfg.linearize(linearizationPoint);
+
+  EXPECT_LONGS_EQUAL(gfg.size(), 3);
+
+  gfg.resize(0);
+  EXPECT_LONGS_EQUAL(gfg.size(), 0);
+}
+
+/*****************************************************************************
+ * Test push_back on HFG makes the correct distinction.
+ */
+TEST(HybridFactorGraph, PushBack) {
+  HybridNonlinearFactorGraph fg;
+
+  auto nonlinearFactor = boost::make_shared<BetweenFactor<double>>();
+  fg.push_back(nonlinearFactor);
+
+  EXPECT_LONGS_EQUAL(fg.size(), 1);
+
+  fg = HybridNonlinearFactorGraph();
+
+  auto discreteFactor = boost::make_shared<DecisionTreeFactor>();
+  fg.push_back(discreteFactor);
+
+  EXPECT_LONGS_EQUAL(fg.size(), 1);
+
+  fg = HybridNonlinearFactorGraph();
+
+  auto dcFactor = boost::make_shared<MixtureFactor>();
+  fg.push_back(dcFactor);
+
+  EXPECT_LONGS_EQUAL(fg.size(), 1);
+
+  // Now do the same with HybridGaussianFactorGraph
+  HybridGaussianFactorGraph ghfg;
+
+  auto gaussianFactor = boost::make_shared<JacobianFactor>();
+  ghfg.push_back(gaussianFactor);
+
+  EXPECT_LONGS_EQUAL(ghfg.size(), 1);
+
+  ghfg = HybridGaussianFactorGraph();
+  ghfg.push_back(discreteFactor);
+
+  EXPECT_LONGS_EQUAL(ghfg.size(), 1);
+
+  ghfg = HybridGaussianFactorGraph();
+  ghfg.push_back(dcFactor);
+
+  EXPECT_LONGS_EQUAL(ghfg.size(), 1);
+}
+
+/****************************************************************************
+ * Test construction of switching-like hybrid factor graph.
+ */
+TEST(HybridFactorGraph, Switching) {
+  Switching self(3);
+
+  EXPECT_LONGS_EQUAL(7, self.nonlinearFactorGraph.size());
+  EXPECT_LONGS_EQUAL(7, self.linearizedFactorGraph.size());
+}
+
+/****************************************************************************
+ * Test linearization on a switching-like hybrid factor graph.
+ */
+TEST(HybridFactorGraph, Linearization) {
+  Switching self(3);
+
+  // Linearize here:
+  HybridGaussianFactorGraph actualLinearized =
+      self.nonlinearFactorGraph.linearize(self.linearizationPoint);
+
+  EXPECT_LONGS_EQUAL(7, actualLinearized.size());
+}
+
+/****************************************************************************
+ * Test elimination tree construction
+ */
+TEST(HybridFactorGraph, EliminationTree) {
+  Switching self(3);
+
+  // Create ordering.
+  Ordering ordering;
+  for (size_t k = 1; k <= self.K; k++) ordering += X(k);
+
+  // Create elimination tree.
+  HybridEliminationTree etree(self.linearizedFactorGraph, ordering);
+  EXPECT_LONGS_EQUAL(1, etree.roots().size())
+}
+
+/****************************************************************************
+ *Test elimination function by eliminating x1 in *-x1-*-x2 graph.
+ */
+TEST(GaussianElimination, Eliminate_x1) {
+  Switching self(3);
+
+  // Gather factors on x1, has a simple Gaussian and a mixture factor.
+  HybridGaussianFactorGraph factors;
+  // Add gaussian prior
+  factors.push_back(self.linearizedFactorGraph[0]);
+  // Add first hybrid factor
+  factors.push_back(self.linearizedFactorGraph[1]);
+
+  // TODO(Varun) remove this block since sum is no longer exposed.
+  // // Check that sum works:
+  // auto sum = factors.sum();
+  // Assignment<Key> mode;
+  // mode[M(1)] = 1;
+  // auto actual = sum(mode);               // Selects one of 2 modes.
+  // EXPECT_LONGS_EQUAL(2, actual.size());  // Prior and motion model.
+
+  // Eliminate x1
+  Ordering ordering;
+  ordering += X(1);
+
+  auto result = EliminateHybrid(factors, ordering);
+  CHECK(result.first);
+  EXPECT_LONGS_EQUAL(1, result.first->nrFrontals());
+  CHECK(result.second);
+  // Has two keys, x2 and m1
+  EXPECT_LONGS_EQUAL(2, result.second->size());
+}
+
+/****************************************************************************
+ * Test elimination function by eliminating x2 in x1-*-x2-*-x3 chain.
+ *                                                m1/      \m2
+ */
+TEST(HybridsGaussianElimination, Eliminate_x2) {
+  Switching self(3);
+
+  // Gather factors on x2, will be two mixture factors (with x1 and x3, resp.).
+  HybridGaussianFactorGraph factors;
+  factors.push_back(self.linearizedFactorGraph[1]);  // involves m1
+  factors.push_back(self.linearizedFactorGraph[2]);  // involves m2
+
+  // TODO(Varun) remove this block since sum is no longer exposed.
+  // // Check that sum works:
+  // auto sum = factors.sum();
+  // Assignment<Key> mode;
+  // mode[M(1)] = 0;
+  // mode[M(2)] = 1;
+  // auto actual = sum(mode);               // Selects one of 4 mode
+  // combinations. EXPECT_LONGS_EQUAL(2, actual.size());  // 2 motion models.
+
+  // Eliminate x2
+  Ordering ordering;
+  ordering += X(2);
+
+  std::pair<HybridConditional::shared_ptr, HybridFactor::shared_ptr> result =
+      EliminateHybrid(factors, ordering);
+  CHECK(result.first);
+  EXPECT_LONGS_EQUAL(1, result.first->nrFrontals());
+  CHECK(result.second);
+  // Note: separator keys should include m1, m2.
+  EXPECT_LONGS_EQUAL(4, result.second->size());
+}
+
+/****************************************************************************
+ * Helper method to generate gaussian factor graphs with a specific mode.
+ */
+GaussianFactorGraph::shared_ptr batchGFG(double between,
+                                         Values linearizationPoint) {
+  NonlinearFactorGraph graph;
+  graph.addPrior<double>(X(1), 0, Isotropic::Sigma(1, 0.1));
+
+  auto between_x1_x2 = boost::make_shared<MotionModel>(
+      X(1), X(2), between, Isotropic::Sigma(1, 1.0));
+
+  graph.push_back(between_x1_x2);
+
+  return graph.linearize(linearizationPoint);
+}
+
+/****************************************************************************
+ * Test elimination function by eliminating x1 and x2 in graph.
+ */
+TEST(HybridGaussianElimination, EliminateHybrid_2_Variable) {
+  Switching self(2, 1.0, 0.1);
+
+  auto factors = self.linearizedFactorGraph;
+
+  // Eliminate x1
+  Ordering ordering;
+  ordering += X(1);
+  ordering += X(2);
+
+  HybridConditional::shared_ptr hybridConditionalMixture;
+  HybridFactor::shared_ptr factorOnModes;
+
+  std::tie(hybridConditionalMixture, factorOnModes) =
+      EliminateHybrid(factors, ordering);
+
+  auto gaussianConditionalMixture =
+      dynamic_pointer_cast<GaussianMixture>(hybridConditionalMixture->inner());
+
+  CHECK(gaussianConditionalMixture);
+  // Frontals = [x1, x2]
+  EXPECT_LONGS_EQUAL(2, gaussianConditionalMixture->nrFrontals());
+  // 1 parent, which is the mode
+  EXPECT_LONGS_EQUAL(1, gaussianConditionalMixture->nrParents());
+
+  // This is now a HybridDiscreteFactor
+  auto hybridDiscreteFactor =
+      dynamic_pointer_cast<HybridDiscreteFactor>(factorOnModes);
+  // Access the type-erased inner object and convert to DecisionTreeFactor
+  auto discreteFactor =
+      dynamic_pointer_cast<DecisionTreeFactor>(hybridDiscreteFactor->inner());
+  CHECK(discreteFactor);
+  EXPECT_LONGS_EQUAL(1, discreteFactor->discreteKeys().size());
+  EXPECT(discreteFactor->root_->isLeaf() == false);
+}
+
+// /*
+// ****************************************************************************/
+// /// Test the toDecisionTreeFactor method
+// TEST(HybridFactorGraph, ToDecisionTreeFactor) {
+//   size_t K = 3;
+
+//   // Provide tight sigma values so that the errors are visibly different.
+//   double between_sigma = 5e-8, prior_sigma = 1e-7;
+
+//   Switching self(K, between_sigma, prior_sigma);
+
+//   // Clear out discrete factors since sum() cannot hanldle those
+//   HybridGaussianFactorGraph linearizedFactorGraph(
+//       self.linearizedFactorGraph.gaussianGraph(), DiscreteFactorGraph(),
+//       self.linearizedFactorGraph.dcGraph());
+
+//   auto decisionTreeFactor = linearizedFactorGraph.toDecisionTreeFactor();
+
+//   auto allAssignments =
+//       DiscreteValues::CartesianProduct(linearizedFactorGraph.discreteKeys());
+
+//   // Get the error of the discrete assignment m1=0, m2=1.
+//   double actual = (*decisionTreeFactor)(allAssignments[1]);
+
+//   /********************************************/
+//   // Create equivalent factor graph for m1=0, m2=1
+//   GaussianFactorGraph graph = linearizedFactorGraph.gaussianGraph();
+
+//   for (auto &p : linearizedFactorGraph.dcGraph()) {
+//     if (auto mixture =
+//             boost::dynamic_pointer_cast<DCGaussianMixtureFactor>(p)) {
+//       graph.add((*mixture)(allAssignments[1]));
+//     }
+//   }
+
+//   VectorValues values = graph.optimize();
+//   double expected = graph.probPrime(values);
+//   /********************************************/
+//   EXPECT_DOUBLES_EQUAL(expected, actual, 1e-12);
+//   // REGRESSION:
+//   EXPECT_DOUBLES_EQUAL(0.6125, actual, 1e-4);
+// }
+
+/****************************************************************************
+ * Test partial elimination
+ */
+TEST(HybridFactorGraph, Partial_Elimination) {
+  Switching self(3);
+
+  auto linearizedFactorGraph = self.linearizedFactorGraph;
+
+  // Create ordering.
+  Ordering ordering;
+  for (size_t k = 1; k <= self.K; k++) ordering += X(k);
+
+  // Eliminate partially.
+  HybridBayesNet::shared_ptr hybridBayesNet;
+  HybridGaussianFactorGraph::shared_ptr remainingFactorGraph;
+  std::tie(hybridBayesNet, remainingFactorGraph) =
+      linearizedFactorGraph.eliminatePartialSequential(ordering);
+
+  CHECK(hybridBayesNet);
+  //  GTSAM_PRINT(*hybridBayesNet);  // HybridBayesNet
+  EXPECT_LONGS_EQUAL(3, hybridBayesNet->size());
+  EXPECT(hybridBayesNet->at(0)->frontals() == KeyVector{X(1)});
+  EXPECT(hybridBayesNet->at(0)->parents() == KeyVector({X(2), M(1)}));
+  EXPECT(hybridBayesNet->at(1)->frontals() == KeyVector{X(2)});
+  EXPECT(hybridBayesNet->at(1)->parents() == KeyVector({X(3), M(1), M(2)}));
+  EXPECT(hybridBayesNet->at(2)->frontals() == KeyVector{X(3)});
+  EXPECT(hybridBayesNet->at(2)->parents() == KeyVector({M(1), M(2)}));
+
+  CHECK(remainingFactorGraph);
+  //  GTSAM_PRINT(*remainingFactorGraph);  // HybridFactorGraph
+  EXPECT_LONGS_EQUAL(3, remainingFactorGraph->size());
+  EXPECT(remainingFactorGraph->at(0)->keys() == KeyVector({M(1)}));
+  EXPECT(remainingFactorGraph->at(1)->keys() == KeyVector({M(2), M(1)}));
+  EXPECT(remainingFactorGraph->at(2)->keys() == KeyVector({M(1), M(2)}));
+}
+
+/****************************************************************************
+ * Test full elimination
+ */
+TEST(HybridFactorGraph, Full_Elimination) {
+  Switching self(3);
+
+  auto linearizedFactorGraph = self.linearizedFactorGraph;
+
+  // We first do a partial elimination
+  HybridBayesNet::shared_ptr hybridBayesNet_partial;
+  HybridGaussianFactorGraph::shared_ptr remainingFactorGraph_partial;
+  DiscreteBayesNet discreteBayesNet;
+
+  {
+    // Create ordering.
+    Ordering ordering;
+    for (size_t k = 1; k <= self.K; k++) ordering += X(k);
+
+    // Eliminate partially.
+    std::tie(hybridBayesNet_partial, remainingFactorGraph_partial) =
+        linearizedFactorGraph.eliminatePartialSequential(ordering);
+
+    DiscreteFactorGraph discrete_fg;
+    // TODO(Varun) Make this a function of HybridGaussianFactorGraph?
+    for (HybridFactor::shared_ptr& factor : (*remainingFactorGraph_partial)) {
+      auto df = dynamic_pointer_cast<HybridDiscreteFactor>(factor);
+      discrete_fg.push_back(df->inner());
+    }
+    ordering.clear();
+    for (size_t k = 1; k < self.K; k++) ordering += M(k);
+    discreteBayesNet =
+        *discrete_fg.eliminateSequential(ordering, EliminateForMPE);
+  }
+
+  // Create ordering.
+  Ordering ordering;
+  for (size_t k = 1; k <= self.K; k++) ordering += X(k);
+  for (size_t k = 1; k < self.K; k++) ordering += M(k);
+
+  // Eliminate partially.
+  HybridBayesNet::shared_ptr hybridBayesNet =
+      linearizedFactorGraph.eliminateSequential(ordering);
+
+  CHECK(hybridBayesNet);
+  EXPECT_LONGS_EQUAL(5, hybridBayesNet->size());
+  // p(x1 | x2, m1)
+  EXPECT(hybridBayesNet->at(0)->frontals() == KeyVector{X(1)});
+  EXPECT(hybridBayesNet->at(0)->parents() == KeyVector({X(2), M(1)}));
+  // p(x2 | x3, m1, m2)
+  EXPECT(hybridBayesNet->at(1)->frontals() == KeyVector{X(2)});
+  EXPECT(hybridBayesNet->at(1)->parents() == KeyVector({X(3), M(1), M(2)}));
+  // p(x3 | m1, m2)
+  EXPECT(hybridBayesNet->at(2)->frontals() == KeyVector{X(3)});
+  EXPECT(hybridBayesNet->at(2)->parents() == KeyVector({M(1), M(2)}));
+  // P(m1 | m2)
+  EXPECT(hybridBayesNet->at(3)->frontals() == KeyVector{M(1)});
+  EXPECT(hybridBayesNet->at(3)->parents() == KeyVector({M(2)}));
+  EXPECT(
+      dynamic_pointer_cast<DiscreteConditional>(hybridBayesNet->at(3)->inner())
+          ->equals(*discreteBayesNet.at(0)));
+  // P(m2)
+  EXPECT(hybridBayesNet->at(4)->frontals() == KeyVector{M(2)});
+  EXPECT_LONGS_EQUAL(0, hybridBayesNet->at(4)->nrParents());
+  EXPECT(
+      dynamic_pointer_cast<DiscreteConditional>(hybridBayesNet->at(4)->inner())
+          ->equals(*discreteBayesNet.at(1)));
+}
+
+/****************************************************************************
+ * Test printing
+ */
+TEST(HybridFactorGraph, Printing) {
+  Switching self(3);
+
+  auto linearizedFactorGraph = self.linearizedFactorGraph;
+
+  // Create ordering.
+  Ordering ordering;
+  for (size_t k = 1; k <= self.K; k++) ordering += X(k);
+
+  // Eliminate partially.
+  HybridBayesNet::shared_ptr hybridBayesNet;
+  HybridGaussianFactorGraph::shared_ptr remainingFactorGraph;
+  std::tie(hybridBayesNet, remainingFactorGraph) =
+      linearizedFactorGraph.eliminatePartialSequential(ordering);
+
+  string expected_hybridFactorGraph = R"(
+size: 7
+factor 0: Continuous [x1]
+
+  A[x1] = [
+	10
+]
+  b = [ -10 ]
+  No noise model
+factor 1: Hybrid [x1 x2; m1]{
+ Choice(m1) 
+ 0 Leaf :
+  A[x1] = [
+	-1
+]
+  A[x2] = [
+	1
+]
+  b = [ -1 ]
+  No noise model
+
+ 1 Leaf :
+  A[x1] = [
+	-1
+]
+  A[x2] = [
+	1
+]
+  b = [ -0 ]
+  No noise model
+
+}
+factor 2: Hybrid [x2 x3; m2]{
+ Choice(m2) 
+ 0 Leaf :
+  A[x2] = [
+	-1
+]
+  A[x3] = [
+	1
+]
+  b = [ -1 ]
+  No noise model
+
+ 1 Leaf :
+  A[x2] = [
+	-1
+]
+  A[x3] = [
+	1
+]
+  b = [ -0 ]
+  No noise model
+
+}
+factor 3: Continuous [x2]
+
+  A[x2] = [
+	10
+]
+  b = [ -10 ]
+  No noise model
+factor 4: Continuous [x3]
+
+  A[x3] = [
+	10
+]
+  b = [ -10 ]
+  No noise model
+factor 5: Discrete [m1]
+ P( m1 ):
+ Leaf  0.5
+
+factor 6: Discrete [m2 m1]
+ P( m2 | m1 ):
+ Choice(m2) 
+ 0 Choice(m1) 
+ 0 0 Leaf 0.33333333
+ 0 1 Leaf  0.6
+ 1 Choice(m1) 
+ 1 0 Leaf 0.66666667
+ 1 1 Leaf  0.4
+
+)";
+  EXPECT(assert_print_equal(expected_hybridFactorGraph, linearizedFactorGraph));
+
+  // Expected output for hybridBayesNet.
+  string expected_hybridBayesNet = R"(
+size: 3
+factor 0: Hybrid  P( x1 | x2 m1)
+ Discrete Keys = (m1, 2), 
+ Choice(m1) 
+ 0 Leaf  p(x1 | x2)
+  R = [ 10.0499 ]
+  S[x2] = [ -0.0995037 ]
+  d = [ -9.85087 ]
+  No noise model
+
+ 1 Leaf  p(x1 | x2)
+  R = [ 10.0499 ]
+  S[x2] = [ -0.0995037 ]
+  d = [ -9.95037 ]
+  No noise model
+
+factor 1: Hybrid  P( x2 | x3 m1 m2)
+ Discrete Keys = (m1, 2), (m2, 2), 
+ Choice(m2) 
+ 0 Choice(m1) 
+ 0 0 Leaf  p(x2 | x3)
+  R = [ 10.099 ]
+  S[x3] = [ -0.0990196 ]
+  d = [ -9.99901 ]
+  No noise model
+
+ 0 1 Leaf  p(x2 | x3)
+  R = [ 10.099 ]
+  S[x3] = [ -0.0990196 ]
+  d = [ -9.90098 ]
+  No noise model
+
+ 1 Choice(m1) 
+ 1 0 Leaf  p(x2 | x3)
+  R = [ 10.099 ]
+  S[x3] = [ -0.0990196 ]
+  d = [ -10.098 ]
+  No noise model
+
+ 1 1 Leaf  p(x2 | x3)
+  R = [ 10.099 ]
+  S[x3] = [ -0.0990196 ]
+  d = [ -10 ]
+  No noise model
+
+factor 2: Hybrid  P( x3 | m1 m2)
+ Discrete Keys = (m1, 2), (m2, 2), 
+ Choice(m2) 
+ 0 Choice(m1) 
+ 0 0 Leaf  p(x3)
+  R = [ 10.0494 ]
+  d = [ -10.1489 ]
+  No noise model
+
+ 0 1 Leaf  p(x3)
+  R = [ 10.0494 ]
+  d = [ -10.1479 ]
+  No noise model
+
+ 1 Choice(m1) 
+ 1 0 Leaf  p(x3)
+  R = [ 10.0494 ]
+  d = [ -10.0504 ]
+  No noise model
+
+ 1 1 Leaf  p(x3)
+  R = [ 10.0494 ]
+  d = [ -10.0494 ]
+  No noise model
+
+)";
+  EXPECT(assert_print_equal(expected_hybridBayesNet, *hybridBayesNet));
+}
+
+/****************************************************************************
+ * Simple PlanarSLAM example test with 2 poses and 2 landmarks (each pose
+ * connects to 1 landmark) to expose issue with default decision tree creation
+ * in hybrid elimination. The hybrid factor is between the poses X0 and X1.
+ * The issue arises if we eliminate a landmark variable first since it is not
+ * connected to a HybridFactor.
+ */
+TEST(HybridFactorGraph, DefaultDecisionTree) {
+  HybridNonlinearFactorGraph fg;
+
+  // Add a prior on pose x1 at the origin.
+  // A prior factor consists of a mean and a noise model (covariance matrix)
+  Pose2 prior(0.0, 0.0, 0.0);  // prior mean is at origin
+  auto priorNoise = noiseModel::Diagonal::Sigmas(
+      Vector3(0.3, 0.3, 0.1));  // 30cm std on x,y, 0.1 rad on theta
+  fg.emplace_nonlinear<PriorFactor<Pose2>>(X(0), prior, priorNoise);
+
+  using PlanarMotionModel = BetweenFactor<Pose2>;
+
+  // Add odometry factor
+  Pose2 odometry(2.0, 0.0, 0.0);
+  KeyVector contKeys = {X(0), X(1)};
+  auto noise_model = noiseModel::Isotropic::Sigma(3, 1.0);
+  auto still = boost::make_shared<PlanarMotionModel>(X(0), X(1), Pose2(0, 0, 0),
+                                                     noise_model),
+       moving = boost::make_shared<PlanarMotionModel>(X(0), X(1), odometry,
+                                                      noise_model);
+  std::vector<PlanarMotionModel::shared_ptr> motion_models = {still, moving};
+  // TODO(Varun) Make a templated constructor for MixtureFactor which does this?
+  std::vector<NonlinearFactor::shared_ptr> components;
+  for (auto&& f : motion_models) {
+    components.push_back(boost::dynamic_pointer_cast<NonlinearFactor>(f));
+  }
+  fg.emplace_hybrid<MixtureFactor>(
+      contKeys, DiscreteKeys{gtsam::DiscreteKey(M(1), 2)}, components);
+
+  // Add Range-Bearing measurements to from X0 to L0 and X1 to L1.
+  // create a noise model for the landmark measurements
+  auto measurementNoise = noiseModel::Diagonal::Sigmas(
+      Vector2(0.1, 0.2));  // 0.1 rad std on bearing, 20cm on range
+  // create the measurement values - indices are (pose id, landmark id)
+  Rot2 bearing11 = Rot2::fromDegrees(45), bearing22 = Rot2::fromDegrees(90);
+  double range11 = std::sqrt(4.0 + 4.0), range22 = 2.0;
+
+  // Add Bearing-Range factors
+  fg.emplace_nonlinear<BearingRangeFactor<Pose2, Point2>>(
+      X(0), L(0), bearing11, range11, measurementNoise);
+  fg.emplace_nonlinear<BearingRangeFactor<Pose2, Point2>>(
+      X(1), L(1), bearing22, range22, measurementNoise);
+
+  // Create (deliberately inaccurate) initial estimate
+  Values initialEstimate;
+  initialEstimate.insert(X(0), Pose2(0.5, 0.0, 0.2));
+  initialEstimate.insert(X(1), Pose2(2.3, 0.1, -0.2));
+  initialEstimate.insert(L(0), Point2(1.8, 2.1));
+  initialEstimate.insert(L(1), Point2(4.1, 1.8));
+
+  // We want to eliminate variables not connected to DCFactors first.
+  Ordering ordering;
+  ordering += L(0);
+  ordering += L(1);
+  ordering += X(0);
+  ordering += X(1);
+
+  HybridGaussianFactorGraph linearized = fg.linearize(initialEstimate);
+  gtsam::HybridBayesNet::shared_ptr hybridBayesNet;
+  gtsam::HybridGaussianFactorGraph::shared_ptr remainingFactorGraph;
+
+  // This should NOT fail
+  std::tie(hybridBayesNet, remainingFactorGraph) =
+      linearized.eliminatePartialSequential(ordering);
+  EXPECT_LONGS_EQUAL(4, hybridBayesNet->size());
+  EXPECT_LONGS_EQUAL(1, remainingFactorGraph->size());
+}
+
+/* ************************************************************************* */
+int main() {
+  TestResult tr;
+  return TestRegistry::runAllTests(tr);
+}
+/* ************************************************************************* */