On macOS double-precision serialization is not lossless, so switch to float.

src/mlpack/methods/cf/normalization/combined_normalization.hpp

@@ -119,7 +119,7 @@ class CombinedNormalization
   void SequenceNormalize(MatType& data)
   {
     std::get<I>(normalizations).Normalize(data);
-    SequenceNormalize<I+1>(data);
+    SequenceNormalize<I + 1>(data);
   }
 
   //! End of tuple unpacking.
@@ -140,7 +140,7 @@ class CombinedNormalization
   {
     // The order of denormalization should be the reversed order
     // of normalization.
-    double realRating = SequenceDenormalize<I+1>(user, item, rating);
+    double realRating = SequenceDenormalize<I + 1>(user, item, rating);
     realRating =
         std::get<I>(normalizations).Denormalize(user, item, realRating);
     return realRating;
@@ -190,7 +190,7 @@ class CombinedNormalization
     tagName += std::to_string(I);
     ar & boost::serialization::make_nvp(
         tagName.c_str(), std::get<I>(normalizations));
-    SequenceSerialize<I+1, Archive>(ar, version);
+    SequenceSerialize<I + 1, Archive>(ar, version);
   }
 
   //! End of tuple unpacking.

src/mlpack/methods/cf/normalization/item_mean_normalization.hpp

@@ -78,7 +78,7 @@ class ItemMeanNormalization
       // The algorithm omits rating of zero. If normalized rating equals zero,
       // it is set to the smallest positive double value.
       if (datapoint(2) == 0)
-        datapoint(2) = std::numeric_limits<double>::min();
+        datapoint(2) = std::numeric_limits<float>::min();
     });
   }
 
@@ -114,7 +114,7 @@ class ItemMeanNormalization
       // The algorithm omits rating of zero. If normalized rating equals zero,
       // it is set to the smallest positive double value.
       if (tmp == 0)
-        tmp = std::numeric_limits<double>::min();
+        tmp = std::numeric_limits<float>::min();
 
       *it = tmp;
     }

src/mlpack/methods/cf/normalization/user_mean_normalization.hpp

@@ -114,7 +114,7 @@ class UserMeanNormalization
       // The algorithm omits rating of zero. If normalized rating equals zero,
       // it is set to the smallest positive double value.
       if (tmp == 0)
-        tmp = std::numeric_limits<double>::min();
+        tmp = std::numeric_limits<float>::min();
 
       *it = tmp;
     }

src/mlpack/methods/cf/normalization/z_score_normalization.hpp

@@ -64,7 +64,7 @@ class ZScoreNormalization
     data.row(2).for_each([](double& x)
     {
       if (x == 0)
-        x = std::numeric_limits<double>::min();
+        x = std::numeric_limits<float>::min();
     });
   }
 
@@ -99,7 +99,7 @@ class ZScoreNormalization
       // The algorithm omits rating of zero. If normalized rating equals zero,
       // it is set to the smallest positive double value.
       if (tmp == 0)
-        tmp = std::numeric_limits<double>::min();
+        tmp = std::numeric_limits<float>::min();
 
       *it = tmp;
     }

src/mlpack/tests/cf_test.cpp

@@ -702,7 +702,7 @@ BOOST_AUTO_TEST_CASE(RecommendationAccuracySVDCompleteTest)
 /**
  * Make sure recommendations that are generated are reasonably accurate
  * for SVD Incomplete Incremental method.
- */ 
+ */
 BOOST_AUTO_TEST_CASE(RecommendationAccuracySVDIncompleteTest)
 {
   RecommendationAccuracy<SVDIncompletePolicy>();
@@ -711,7 +711,7 @@ BOOST_AUTO_TEST_CASE(RecommendationAccuracySVDIncompleteTest)
 /**
  * Make sure recommendations that are generated are reasonably accurate
  * for Bias SVD method.
- */ 
+ */
 BOOST_AUTO_TEST_CASE(RecommendationAccuracyBiasSVDTest)
 {
   RecommendationAccuracy<BiasSVDPolicy>();