More Efficient Nudge In Progess

include/random_walks/gaussian_hamiltonian_monte_carlo_exact_walk.hpp

@@ -241,13 +241,16 @@ private :
     {
         MT A = P.get_mat();
         VT b = P.get_vec();
-
         int m = A.rows();
-        const NT* b_data = b.data();
+
+        VT b_Ax = b - A * p.getCoefficients();
+        const NT* b_Ax_data = b_Ax.data();
+
+        NT dist;
 
         for (int i = 0; i < m; i++) {
 
-            NT dist = *b_data - A.row(i) * p.getCoefficients();
+            dist = *b_Ax_data;
 
             if (dist < NT(-tol)){
                 //Nudging correction
@@ -263,7 +266,7 @@ private :
                 shift.operator*=(eps_2);
                 p.operator+=(shift);
             }
-            b_data++;
+            b_Ax_data++;
         }
     }
 

test/gaussian_hmc_reflections_test.cpp

@@ -0,0 +1,87 @@
+#include "Eigen/Eigen"
+#include <iostream>
+#include "random.hpp"
+#include "random/uniform_int.hpp"
+#include "random/normal_distribution.hpp"
+#include "cartesian_geom/cartesian_kernel.h"
+#include "convex_bodies/hpolytope.h"
+#include "generators/known_polytope_generators.h"
+#include "generators/h_polytopes_generator.h" // Include your polytope generator
+#include "random_walks/random_walks.hpp"
+#include "misc/misc.h"
+#include "doctest.h"
+
+typedef double NT;
+typedef Cartesian<NT> Kernel;
+typedef typename Kernel::Point Point;
+typedef BoostRandomNumberGenerator<boost::mt19937, NT, 3> RNGType;
+typedef HPolytope<Point> Polytope;
+
+void test_gaussian_hmc_reflections(const std::string& polytope_type, unsigned int walk_length) {
+    Polytope P;
+    unsigned int dim = 3; //dimension
+    if (polytope_type == "cube") 
+    {
+        std::cout<<std::endl<<"CUBE"<<std::endl<<std::endl;
+        P = generate_cube<Polytope>(dim, false);
+    } 
+    else if (polytope_type == "cross") 
+    {
+        P = generate_cross<Polytope>(dim, false);
+    }
+    else if (polytope_type == "simplex") 
+    {
+        P = generate_simplex<Polytope>(dim, false);
+    } 
+    else if (polytope_type == "randomHPoly") 
+    {
+        std::cout<<std::endl<<"RANDOM HPPOLY"<<std::endl<<std::endl;
+        unsigned int m = 100; //number of hyperplanes
+        P = random_hpoly<Polytope, boost::mt19937>(dim, m, 123456);
+    }
+    else {
+        std::cout << "Unknown polytope type: " << polytope_type << std::endl;
+        return;
+    }
+
+    Point interior_point = P.ComputeInnerBall().first;
+    double L = 100.0; // Step size
+    unsigned int rho = 100; // Max reflections
+    RNGType rng(123456);
+    //typename GaussianHamiltonianMonteCarloExactWalk::parameters params(L, true, rho, true);
+    //typename GaussianHamiltonianMonteCarloExactWalk::Walk<Polytope, RNGType> walk(P, interior_point, 1.0, rng, params);
+    typename GaussianHamiltonianMonteCarloExactWalk::Walk<Polytope, RNGType> walk(P, interior_point, 1.0, rng);
+    walk.apply(P, interior_point, 1.0, walk_length, rng);
+}
+
+
+
+
+
+
+
+
+
+TEST_CASE("GHMC Reflections Test - Cube") {
+    unsigned int walk_length = 100;
+    test_gaussian_hmc_reflections("cube", walk_length);
+}
+
+
+
+TEST_CASE("GHMC Reflections Test - Random HPolytope") {
+    RNGType rng(123456);
+    test_gaussian_hmc_reflections("randomHPoly", 10); // 3D polytope with 30 hyperplanes
+}
+
+
+TEST_CASE("GHMC Reflections Test - Simplex") {
+    RNGType rng(123456);
+    test_gaussian_hmc_reflections("simplex", 10);
+}
+
+
+TEST_CASE("GHMC Reflections Test - Cross Polytope") {
+    RNGType rng(1234567);
+    test_gaussian_hmc_reflections("cross", 1000);
+}