MultiGP

docs/tutorials/gp.rst

@@ -13,7 +13,7 @@ We assume that our samples are in a vector called ``samples`` and that our obser
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 79-88
+   :lines: 77-86
 
 Basic usage
 ------------
@@ -23,14 +23,14 @@ We first create a basic GP with an Exponential kernel (``kernel::Exp<Params>``)
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 61-74
+   :lines: 61-72
 
 The type of the GP is defined by the following lines:
 
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 89-93
+   :lines: 87-91
 
 To use the GP, we need :
 
@@ -40,7 +40,7 @@ To use the GP, we need :
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 94-99
+   :lines: 92-97
 
 Here we assume that the noise is the same for all samples and that it is equal to 0.01.
 
@@ -57,7 +57,7 @@ To visualize the predictions of the GP, we can query it for many points and reco
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 101-112
+   :lines: 99-110
 
 
 Hyper-parameter optimization
@@ -71,7 +71,7 @@ A new GP type is defined as follows:
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 114-118
+   :lines: 112-116
 
 It uses the default values for the parameters of ``SquaredExpARD``:
 
@@ -85,7 +85,7 @@ After calling the ``compute()`` method, the hyper-parameters can be optimized by
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 121-123
+   :lines: 119-121
 
 
 We can have a look at the difference between the two GPs:
@@ -115,7 +115,7 @@ We can also save our optimized GP model:
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 140-141
+   :lines: 138-139
 
 This will create a directory called ``myGP`` with several files (the GP data, kernel hyperparameters etc.). If we want a binary format (i.e., more compact), we can replace the ``TextArchive`` by ``BinaryArchive``.
 
@@ -124,6 +124,6 @@ To the load a saved model, we can do the following:
 .. literalinclude:: ../../src/tutorials/gp.cpp
    :language: c++
    :linenos:
-   :lines: 143-144
+   :lines: 141-142
 
 Note that we need to have the same kernel and mean function (i.e., the same GP type) as the one used for saving.

src/examples/mono_dim.cpp

@@ -105,9 +105,6 @@ BO_PARAMS(std::cout,
 
               struct opt_rprop : public defaults::opt_rprop {
               };
-
-              struct opt_parallelrepeater : public defaults::opt_parallelrepeater {
-              };
           };)
 
 struct fit_eval {

src/examples/obs_multi_auto_mean.cpp

@@ -93,9 +93,6 @@ struct Params {
     struct stop_maxiterations {
         BO_PARAM(int, iterations, 100);
     };
-
-    struct opt_parallelrepeater : defaults::opt_parallelrepeater {
-    };
 };
 
 template <typename Params, typename Model>

src/limbo/model.hpp

@@ -50,6 +50,7 @@
 ///@defgroup model_opt_defaults
 
 #include <limbo/model/gp.hpp>
+#include <limbo/model/multi_gp.hpp>
 #include <limbo/model/sparsified_gp.hpp>
 
 #include <limbo/model/gp/kernel_lf_opt.hpp>

src/limbo/model/gp.hpp

@@ -80,7 +80,7 @@ namespace limbo {
             /// useful because the model might be created before knowing anything about the process
             GP() : _dim_in(-1), _dim_out(-1), _inv_kernel_updated(false) {}
 
-            /// useful because the model might be created  before having samples
+            /// useful because the model might be created before having samples
             GP(int dim_in, int dim_out)
                 : _dim_in(dim_in), _dim_out(dim_out), _kernel_function(dim_in), _mean_function(dim_out), _inv_kernel_updated(false) {}
 
@@ -153,7 +153,7 @@ namespace limbo {
 
             /**
              \\rst
-             return :math:`\mu`, :math:`\sigma^2` (unormalized). If there is no sample, return the value according to the mean function. Using this method instead of separate calls to mu() and sigma() is more efficient because some computations are shared between mu() and sigma().
+             return :math:`\mu`, :math:`\sigma^2` (un-normalized). If there is no sample, return the value according to the mean function. Using this method instead of separate calls to mu() and sigma() is more efficient because some computations are shared between mu() and sigma().
              \\endrst
             */
             std::tuple<Eigen::VectorXd, double> query(const Eigen::VectorXd& v) const
@@ -193,14 +193,14 @@ namespace limbo {
             /// return the number of dimensions of the input
             int dim_in() const
             {
-                assert(_dim_in != -1); // need to compute first !
+                assert(_dim_in != -1); // need to compute first!
                 return _dim_in;
             }
 
             /// return the number of dimensions of the output
             int dim_out() const
             {
-                assert(_dim_out != -1); // need to compute first !
+                assert(_dim_out != -1); // need to compute first!
                 return _dim_out;
             }
 

src/limbo/model/gp/hp_opt.hpp

@@ -48,15 +48,14 @@
 
 #include <Eigen/Core>
 
-#include <limbo/opt/parallel_repeater.hpp>
 #include <limbo/opt/rprop.hpp>
 
 namespace limbo {
     namespace model {
         namespace gp {
             ///@ingroup model_opt
             ///base class for optimization of the hyper-parameters of a GP
-            template <typename Params, typename Optimizer = opt::ParallelRepeater<Params, opt::Rprop<Params>>>
+            template <typename Params, typename Optimizer = opt::Rprop<Params>>
             struct HPOpt {
             public:
                 HPOpt() : _called(false) {}

src/limbo/model/gp/kernel_lf_opt.hpp

@@ -47,14 +47,13 @@
 #define LIMBO_MODEL_GP_KERNEL_LF_OPT_HPP
 
 #include <limbo/model/gp/hp_opt.hpp>
-#include <limbo/tools/random_generator.hpp>
 
 namespace limbo {
     namespace model {
         namespace gp {
             ///@ingroup model_opt
             ///optimize the likelihood of the kernel only
-            template <typename Params, typename Optimizer = opt::ParallelRepeater<Params, opt::Rprop<Params>>>
+            template <typename Params, typename Optimizer = opt::Rprop<Params>>
             struct KernelLFOpt : public HPOpt<Params, Optimizer> {
             public:
                 template <typename GP>
@@ -96,8 +95,8 @@ namespace limbo {
                     const GP& _original_gp;
                 };
             };
-        }
-    }
-}
+        } // namespace gp
+    } // namespace model
+} // namespace limbo
 
 #endif

src/limbo/model/gp/kernel_loo_opt.hpp

@@ -47,14 +47,13 @@
 #define LIMBO_MODEL_GP_KERNEL_LOO_OPT_HPP
 
 #include <limbo/model/gp/hp_opt.hpp>
-#include <limbo/tools/random_generator.hpp>
 
 namespace limbo {
     namespace model {
         namespace gp {
             ///@ingroup model_opt
             ///optimize the likelihood of the kernel only
-            template <typename Params, typename Optimizer = opt::ParallelRepeater<Params, opt::Rprop<Params>>>
+            template <typename Params, typename Optimizer = opt::Rprop<Params>>
             struct KernelLooOpt : public HPOpt<Params, Optimizer> {
             public:
                 template <typename GP>
@@ -96,8 +95,8 @@ namespace limbo {
                     const GP& _original_gp;
                 };
             };
-        }
-    }
-}
+        } // namespace gp
+    } // namespace model
+} // namespace limbo
 
 #endif

src/limbo/model/gp/kernel_mean_lf_opt.hpp

@@ -47,14 +47,13 @@
 #define LIMBO_MODEL_GP_KERNEL_MEAN_LF_OPT_HPP
 
 #include <limbo/model/gp/hp_opt.hpp>
-#include <limbo/tools/random_generator.hpp>
 
 namespace limbo {
     namespace model {
         namespace gp {
             ///@ingroup model_opt
             ///optimize the likelihood of both the kernel and the mean (try to align the mean function)
-            template <typename Params, typename Optimizer = opt::ParallelRepeater<Params, opt::Rprop<Params>>>
+            template <typename Params, typename Optimizer = opt::Rprop<Params>>
             struct KernelMeanLFOpt : public HPOpt<Params, Optimizer> {
             public:
                 template <typename GP>
@@ -109,8 +108,8 @@ namespace limbo {
                     const GP& _original_gp;
                 };
             };
-        }
-    }
-}
+        } // namespace gp
+    } // namespace model
+} // namespace limbo
 
 #endif

src/limbo/model/gp/mean_lf_opt.hpp

@@ -47,14 +47,13 @@
 #define LIMBO_MODEL_GP_MEAN_LF_OPT_HPP
 
 #include <limbo/model/gp/hp_opt.hpp>
-#include <limbo/tools/random_generator.hpp>
 
 namespace limbo {
     namespace model {
         namespace gp {
             ///@ingroup model_opt
             ///optimize the likelihood of the mean only (try to align the mean function)
-            template <typename Params, typename Optimizer = opt::ParallelRepeater<Params, opt::Rprop<Params>>>
+            template <typename Params, typename Optimizer = opt::Rprop<Params>>
             struct MeanLFOpt : public HPOpt<Params, Optimizer> {
             public:
                 template <typename GP>
@@ -99,8 +98,8 @@ namespace limbo {
                     GP _original_gp;
                 };
             };
-        }
-    }
-}
+        } // namespace gp
+    } // namespace model
+} // namespace limbo
 
 #endif