Add C++ Roll/Unroll support

setup.py

@@ -21,7 +21,7 @@ include_dirs = [
     xbob.core.get_include(),
     ]
 
-packages = ['bob-io >= 1.2.2', 'bob-machine >= 1.2.2']
+packages = ['bob-io >= 2.0.0a2']
 version = '2.0.0a0'
 
 setup(
@@ -66,6 +66,7 @@ setup(
           "xbob/learn/mlp/rprop.cpp",
           "xbob/learn/mlp/backprop.cpp",
           "xbob/learn/mlp/trainer.cpp",
+          "xbob/learn/mlp/cxx/roll.cpp",
           "xbob/learn/mlp/cxx/machine.cpp",
           "xbob/learn/mlp/cxx/cross_entropy.cpp",
           "xbob/learn/mlp/cxx/square_error.cpp",

xbob/learn/mlp/cxx/roll.cpp

+/**
+ * @author Laurent El Shafey <Laurent.El-Shafey@idiap.ch>
+ * @date Tue Jun 25 18:52:26 CEST 2013
+ *
+ * Copyright (C) 2011-2014 Idiap Research Institute, Martigny, Switzerland
+ */
+
+#include <bob/core/assert.h>
+#include <xbob.learn.mlp/roll.h>
+
+int bob::learn::mlp::detail::getNbParameters(const bob::learn::mlp::Machine& machine)
+{
+  const std::vector<blitz::Array<double,1> >& b = machine.getBiases();
+  const std::vector<blitz::Array<double,2> >& w = machine.getWeights();
+  return bob::learn::mlp::detail::getNbParameters(w, b);
+}
+
+int bob::learn::mlp::detail::getNbParameters(
+  const std::vector<blitz::Array<double,2> >& w,
+  const std::vector<blitz::Array<double,1> >& b)
+{
+  bob::core::array::assertSameDimensionLength(w.size(), b.size());
+  int N = 0;
+  for (int i=0; i<(int)w.size(); ++i)
+    N += b[i].numElements() + w[i].numElements();
+  return N;
+}
+
+void bob::learn::mlp::unroll(const bob::learn::mlp::Machine& machine,
+  blitz::Array<double,1>& vec)
+{
+  const std::vector<blitz::Array<double,1> >& b = machine.getBiases();
+  const std::vector<blitz::Array<double,2> >& w = machine.getWeights();
+  unroll(w, b, vec);
+}
+
+void bob::learn::mlp::unroll(const std::vector<blitz::Array<double,2> >& w,
+  const std::vector<blitz::Array<double,1> >& b, blitz::Array<double,1>& vec)
+{
+  // 1/ Check number of elements
+  const int N = bob::learn::mlp::detail::getNbParameters(w, b);
+  bob::core::array::assertSameDimensionLength(vec.extent(0), N);
+
+  // 2/ Roll
+  blitz::Range rall = blitz::Range::all();
+  int offset=0;
+  for (int i=0; i<(int)w.size(); ++i)
+  {
+    const int Nb = b[i].extent(0);
+    blitz::Range rb(offset,offset+Nb-1);
+    vec(rb) = b[i];
+    offset += Nb;
+
+    const int Nw0 = w[i].extent(0);
+    const int Nw1 = w[i].extent(1);
+    blitz::TinyVector<int,1> tv(Nw1);
+    for (int j=0; j<Nw0; ++j)
+    {
+      blitz::Range rw(offset,offset+Nw1-1);
+      vec(rw) = w[i](j,rall);
+      offset += Nw1;
+    }
+  }
+}
+
+void bob::learn::mlp::roll(bob::learn::mlp::Machine& machine,
+  const blitz::Array<double,1>& vec)
+{
+  std::vector<blitz::Array<double,1> >& b = machine.updateBiases();
+  std::vector<blitz::Array<double,2> >& w = machine.updateWeights();
+  roll(w, b, vec);
+}
+
+void bob::learn::mlp::roll(std::vector<blitz::Array<double,2> >& w,
+  std::vector<blitz::Array<double,1> >& b, const blitz::Array<double,1>& vec)
+{
+  // 1/ Check number of elements
+  const int N = bob::learn::mlp::detail::getNbParameters(w, b);
+  bob::core::array::assertSameDimensionLength(vec.extent(0), N);
+
+  // 2/ Roll
+  blitz::Range rall = blitz::Range::all();
+  int offset=0;
+  for (int i=0; i<(int)w.size(); ++i)
+  {
+    const int Nb = b[i].extent(0);
+    blitz::Array<double,1> vb = vec(blitz::Range(offset,offset+Nb-1));
+    b[i] = vb;
+    offset += Nb;
+
+    const int Nw0 = w[i].extent(0);
+    const int Nw1 = w[i].extent(1);
+    blitz::TinyVector<int,1> tv(Nw1);
+    for (int j=0; j<Nw0; ++j)
+    {
+      blitz::Array<double,1> vw = vec(blitz::Range(offset,offset+Nw1-1));
+      w[i](j,rall) = vw;
+      offset += Nw1;
+    }
+  }
+}

xbob/learn/mlp/include/xbob.learn.mlp/roll.h

+/**
+ * @author Laurent El Shafey <Laurent.El-Shafey@idiap.ch>
+ * @date Tue Jun 25 18:48:20 CEST 2013
+ *
+ * Copyright (C) 2011-2014 Idiap Research Institute, Martigny, Switzerland
+ */
+
+#ifndef BOB_LEARN_MLP_ROLL_H
+#define BOB_LEARN_MLP_ROLL_H
+
+#include <vector>
+#include <blitz/array.h>
+
+#include "machine.h"
+
+namespace bob { namespace learn { namespace mlp {
+
+namespace detail {
+  /**
+   * @brief Returns the number of parameters (weights and biases) in an
+   * MLP.
+   */
+  int getNbParameters(const bob::learn::mlp::Machine& machine);
+  /**
+   * @brief Returns the number of parameters (weights and biases).
+   */
+  int getNbParameters(const std::vector<blitz::Array<double,2> >& weights,
+    const std::vector<blitz::Array<double,1> >& biases);
+}
+
+/**
+ * @brief Puts the parameters (weights and biases) of the machine in a
+ * large single 1D vector
+ */
+void unroll(const bob::learn::mlp::Machine& machine, blitz::Array<double,1>& vec);
+/**
+ * @brief Puts the parameters (weights and biases) in a large single 1D vector
+ */
+void unroll(const std::vector<blitz::Array<double,2> >& weights,
+  const std::vector<blitz::Array<double,1> >& biases,
+  blitz::Array<double,1>& vec);
+
+/**
+ * @brief Sets the parameters (weights and biases) of the machine from a
+ * large single 1D vector
+ */
+void roll(bob::learn::mlp::Machine& machine, const blitz::Array<double,1>& vec);
+
+/**
+ * @brief Sets the parameters (weights and biases) from a
+ * large single 1D vector
+ */
+void roll(std::vector<blitz::Array<double,2> >& weights,
+  std::vector<blitz::Array<double,1> >& biases,
+  const blitz::Array<double,1>& vec);
+
+}}}
+
+#endif /* BOB_LEARN_MLP_ROLL_H */