Added Sparse (L1) criterion

5 files changed, 66 insertions, 0 deletions

diff --git a/todo/SparseCriterion.lua b/SparseCriterion.lua
index 79404a0..ddaa75c 100644
--- a/todo/SparseCriterion.lua
+++ b/SparseCriterion.lua
@@ -5,6 +5,16 @@ function SparseCriterion:__init()
    self.sizeAverage = true
 end
 
+function SparseCriterion:forward(input)
+   input.nn.SparseCriterion_forward(self, input)
+   return self.output
+end
+
+function SparseCriterion:backward(input)
+   input.nn.SparseCriterion_backward(self, input)
+   return self.gradInput
+end
+
 function SparseCriterion:write(file)
    parent.write(self, file)
    file:writeBool(self.sizeAverage)
diff --git a/generic/SparseCriterion.c b/generic/SparseCriterion.c
new file mode 100644
index 0000000..76da569
--- /dev/null
+++ b/generic/SparseCriterion.c
@@ -0,0 +1,49 @@
+#ifndef TH_GENERIC_FILE
+#define TH_GENERIC_FILE "generic/SparseCriterion.c"
+#else
+
+static int nn_(SparseCriterion_forward)(lua_State *L)
+{
+  THTensor *input = luaT_checkudata(L, 2, torch_(Tensor_id));  
+  int sizeAverage = luaT_getfieldcheckboolean(L, 1, "sizeAverage");
+  real sum = 0;
+
+  TH_TENSOR_APPLY(real, input, sum += fabs(*input_data);)
+
+  if(sizeAverage) sum /= THTensor_(nElement)(input);
+
+  lua_pushnumber(L, sum);
+  lua_setfield(L, 1, "output");
+
+  lua_pushnumber(L, sum);
+  return 1;
+}
+
+static int nn_(SparseCriterion_backward)(lua_State *L)
+{
+  THTensor *input = luaT_checkudata(L, 2, torch_(Tensor_id));
+  int sizeAverage = luaT_getfieldcheckboolean(L, 1, "sizeAverage");
+  THTensor *gradInput = luaT_getfieldcheckudata(L, 1, "gradInput", torch_(Tensor_id));
+  real norm = (sizeAverage ? 1./((real)THTensor_(nElement)(input)) : 1.);
+
+  THTensor_(resizeAs)(gradInput, input);
+  TH_TENSOR_APPLY2(real, gradInput, real, input,
+                   *gradInput_data = ( *input_data >= 0 ? norm : -norm);)
+
+  return 1;
+}
+
+static const struct luaL_Reg nn_(SparseCriterion__) [] = {
+  {"SparseCriterion_forward", nn_(SparseCriterion_forward)},
+  {"SparseCriterion_backward", nn_(SparseCriterion_backward)},
+  {NULL, NULL}
+};
+
+static void nn_(SparseCriterion_init)(lua_State *L)
+{
+  luaT_pushmetaclass(L, torch_(Tensor_id));
+  luaT_registeratname(L, nn_(SparseCriterion__), "nn");
+  lua_pop(L,1);
+}
+
+#endif
diff --git a/init.c b/init.c
index 50c5482..b8238c4 100644
--- a/init.c
+++ b/init.c
@@ -33,6 +33,9 @@ static const void* torch_DoubleTensor_id = NULL;
 #include "generic/SpatialReSampling.c"
 #include "THGenerateFloatTypes.h"
 
+#include "generic/SparseCriterion.c"
+#include "THGenerateFloatTypes.h"
+
 #include "generic/Threshold.c"
 #include "THGenerateFloatTypes.h"
 
@@ -50,6 +53,7 @@ DLL_EXPORT int luaopen_libnnx(lua_State *L)
   nn_FloatSpatialMaxPooling_init(L);
   nn_FloatSpatialUpSampling_init(L);
   nn_FloatSpatialReSampling_init(L);
+  nn_FloatSparseCriterion_init(L);
 
   nn_DoubleSpatialLinear_init(L);
   nn_DoubleHardShrink_init(L);
@@ -60,6 +64,7 @@ DLL_EXPORT int luaopen_libnnx(lua_State *L)
   nn_DoubleSpatialMaxPooling_init(L);
   nn_DoubleSpatialUpSampling_init(L);
   nn_DoubleSpatialReSampling_init(L);
+  nn_DoubleSparseCriterion_init(L);
 
   return 1;
 }
diff --git a/init.lua b/init.lua
index cc53e7e..7d92558 100644
--- a/init.lua
+++ b/init.lua
@@ -82,6 +82,7 @@ torch.include('nnx', 'SpatialFovea.lua')
 
 -- criterions:
 torch.include('nnx', 'SuperCriterion.lua')
+torch.include('nnx', 'SparseCriterion.lua')
 torch.include('nnx', 'SpatialMSECriterion.lua')
 torch.include('nnx', 'SpatialClassNLLCriterion.lua')
 
diff --git a/nnx-1.0-1.rockspec b/nnx-1.0-1.rockspec
index 1073da2..7026d4d 100644
--- a/nnx-1.0-1.rockspec
+++ b/nnx-1.0-1.rockspec
@@ -86,6 +86,7 @@ build = {
          install_files(/lua/nnx SpatialFovea.lua)
          install_files(/lua/nnx SpatialMSECriterion.lua)
          install_files(/lua/nnx SpatialClassNLLCriterion.lua)
+         install_files(/lua/nnx SparseCriterion.lua)
          add_subdirectory (test)
          install_targets(/lib nnx)
    ]],