Added test for multi-gp recompute

resibots · Jun 25, 2018 · 1850c5c · 1850c5c
1 parent 6c82699
commit 1850c5c
Show file tree

Hide file tree

Showing 2 changed files with 44 additions and 3 deletions.
diff --git a/src/limbo/model/multi_gp.hpp b/src/limbo/model/multi_gp.hpp
@@ -54,8 +54,8 @@ namespace limbo {
         /// A wrapper for N-output Gaussian processes.
         /// It is parametrized by:
         /// - GP class
-        /// - a kernel function
-        /// - a mean function
+        /// - a kernel function (the same type for all GPs, but can have different parameters)
+        /// - a mean function (the same type and parameters for all GPs)
         /// - [optional] an optimizer for the hyper-parameters
         template <typename Params, template <typename, typename, typename, typename> class GPClass, typename KernelFunction, typename MeanFunction, class HyperParamsOptimizer = limbo::model::gp::NoLFOpt<Params>>
         class MultiGP {
@@ -69,7 +69,6 @@ namespace limbo {
             MultiGP(int dim_in, int dim_out)
                 : _dim_in(dim_in), _dim_out(dim_out), _mean_function(dim_out)
             {
-
                 // initialize dim_in models with 1 output
                 _gp_models.resize(_dim_out);
                 for (int i = 0; i < _dim_out; i++) {

diff --git a/src/tests/test_gp.cpp b/src/tests/test_gp.cpp
@@ -1112,3 +1112,45 @@ BOOST_AUTO_TEST_CASE(test_multi_gp_auto)
     BOOST_CHECK(sigma(0) <= 2. * (gp.gp_models()[0].kernel_function().noise() + 1e-8));
     BOOST_CHECK(sigma(1) <= 2. * (gp.gp_models()[1].kernel_function().noise() + 1e-8));
 }
+
+BOOST_AUTO_TEST_CASE(test_multi_gp_recompute)
+{
+    using KF_t = kernel::SquaredExpARD<Params>;
+    using Mean_t = mean::Constant<Params>;
+    using MultiGP_t = model::MultiGP<Params, model::GP, KF_t, Mean_t>;
+
+    MultiGP_t gp;
+
+    gp.add_sample(make_v2(1, 1), make_v1(2));
+    gp.add_sample(make_v2(2, 2), make_v1(10));
+
+    // make sure that the observations are properly passed
+    BOOST_CHECK(gp._observations[0](0) == 2.);
+    BOOST_CHECK(gp._observations[1](0) == 10.);
+
+    // make sure the child GPs have the proper observations
+    BOOST_CHECK(gp.gp_models()[0]._observations.row(0)[0] == (2. - gp.mean_function().h_params()[0]));
+    BOOST_CHECK(gp.gp_models()[0]._observations.row(1)[0] == (10. - gp.mean_function().h_params()[0]));
+
+    // now change the mean function parameters
+    gp.mean_function().set_h_params(make_v1(2));
+
+    // make sure that the observations are properly passed
+    BOOST_CHECK(gp._observations[0](0) == 2.);
+    BOOST_CHECK(gp._observations[1](0) == 10.);
+
+    // make sure the child GPs do not have the proper observations
+    BOOST_CHECK(!(gp.gp_models()[0]._observations.row(0)[0] == (2. - gp.mean_function().h_params()[0])));
+    BOOST_CHECK(!(gp.gp_models()[0]._observations.row(1)[0] == (10. - gp.mean_function().h_params()[0])));
+
+    // recompute the GP
+    gp.recompute();
+
+    // make sure that the observations are properly passed
+    BOOST_CHECK(gp._observations[0](0) == 2.);
+    BOOST_CHECK(gp._observations[1](0) == 10.);
+
+    // make sure the child GPs have the proper observations
+    BOOST_CHECK(gp.gp_models()[0]._observations.row(0)[0] == (2. - gp.mean_function().h_params()[0]));
+    BOOST_CHECK(gp.gp_models()[0]._observations.row(1)[0] == (10. - gp.mean_function().h_params()[0]));
+}