Max bin by feature (#2190)

* Add parameter max_bin_by_feature.

* Fix minor bug.

* Fix minor bug.

* Fix calculation of header size for writing binary file.

* Fix style issues.

* Fix python style issue.

* Fix test and python style issue.

docs/Parameters.rst

@@ -447,6 +447,12 @@ IO Parameters
 
    -  LightGBM will auto compress memory according to ``max_bin``. For example, LightGBM will use ``uint8_t`` for feature value if ``max_bin=255``
 
+-  ``max_bin_by_feature`` :raw-html:`<a id="max_bin_by_feature" title="Permalink to this parameter" href="#max_bin_by_feature">&#x1F517;&#xFE0E;</a>`, default = ``None``, type = multi-int
+
+   -  max number of bins for each feature
+
+   -  if not specified, will use ``max_bin`` for all features
+
 -  ``min_data_in_bin`` :raw-html:`<a id="min_data_in_bin" title="Permalink to this parameter" href="#min_data_in_bin">&#x1F517;&#xFE0E;</a>`, default = ``3``, type = int, constraints: ``min_data_in_bin > 0``
 
    -  minimal number of data inside one bin

include/LightGBM/config.h

@@ -443,6 +443,12 @@ struct Config {
   // desc = LightGBM will auto compress memory according to ``max_bin``. For example, LightGBM will use ``uint8_t`` for feature value if ``max_bin=255``
   int max_bin = 255;
 
+  // type = multi-int
+  // default = None
+  // desc = max number of bins for each feature
+  // desc = if not specified, will use ``max_bin`` for all features
+  std::vector<int32_t> max_bin_by_feature;  
+
   // check = >0
   // desc = minimal number of data inside one bin
   // desc = use this to avoid one-data-one-bin (potential over-fitting)

include/LightGBM/dataset.h

@@ -624,6 +624,7 @@ class Dataset {
   std::vector<double> feature_penalty_;
   bool is_finish_load_;
   int max_bin_;
+  std::vector<int32_t> max_bin_by_feature_;
   int bin_construct_sample_cnt_;
   int min_data_in_bin_;
   bool use_missing_;

src/io/config_auto.cpp

@@ -218,6 +218,7 @@ std::unordered_set<std::string> Config::parameter_set({
   "cegb_penalty_feature_coupled",
   "verbosity",
   "max_bin",
+  "max_bin_by_feature",
   "min_data_in_bin",
   "bin_construct_sample_cnt",
   "histogram_pool_size",
@@ -418,6 +419,10 @@ void Config::GetMembersFromString(const std::unordered_map<std::string, std::str
   GetInt(params, "max_bin", &max_bin);
   CHECK(max_bin >1);
 
+  if (GetString(params, "max_bin_by_feature", &tmp_str)) {
+    max_bin_by_feature = Common::StringToArray<int32_t>(tmp_str, ',');
+  }
+
   GetInt(params, "min_data_in_bin", &min_data_in_bin);
   CHECK(min_data_in_bin >0);
 
@@ -610,6 +615,7 @@ std::string Config::SaveMembersToString() const {
   str_buf << "[cegb_penalty_feature_coupled: " << Common::Join(cegb_penalty_feature_coupled, ",") << "]\n";
   str_buf << "[verbosity: " << verbosity << "]\n";
   str_buf << "[max_bin: " << max_bin << "]\n";
+  str_buf << "[max_bin_by_feature: " << Common::Join(max_bin_by_feature, ",") << "]\n";
   str_buf << "[min_data_in_bin: " << min_data_in_bin << "]\n";
   str_buf << "[bin_construct_sample_cnt: " << bin_construct_sample_cnt << "]\n";
   str_buf << "[histogram_pool_size: " << histogram_pool_size << "]\n";

src/io/dataset.cpp

@@ -318,6 +318,12 @@ void Dataset::Construct(
       feature_penalty_.clear();
     }
   }
+  if (!io_config.max_bin_by_feature.empty()) {
+    CHECK(static_cast<size_t>(num_total_features_) == io_config.max_bin_by_feature.size());
+    CHECK(*(std::min_element(io_config.max_bin_by_feature.begin(), io_config.max_bin_by_feature.end())) > 1);
+    max_bin_by_feature_.resize(num_total_features_);
+    max_bin_by_feature_.assign(io_config.max_bin_by_feature.begin(), io_config.max_bin_by_feature.end());
+  }
   max_bin_ = io_config.max_bin;
   min_data_in_bin_ = io_config.min_data_in_bin;
   bin_construct_sample_cnt_ = io_config.bin_construct_sample_cnt;
@@ -332,6 +338,9 @@ void Dataset::ResetConfig(const char* parameters) {
   if (param.count("max_bin") && io_config.max_bin != max_bin_) {
     Log::Warning("Cannot change max_bin after constructed Dataset handle.");
   }
+  if (param.count("max_bin_by_feature") && io_config.max_bin_by_feature != max_bin_by_feature_) {
+    Log::Warning("Cannot change max_bin_by_feature after constructed Dataset handle.");
+  }
   if (param.count("bin_construct_sample_cnt") && io_config.bin_construct_sample_cnt != bin_construct_sample_cnt_) {
     Log::Warning("Cannot change bin_construct_sample_cnt after constructed Dataset handle.");
   }
@@ -643,7 +652,7 @@ void Dataset::SaveBinaryFile(const char* bin_filename) {
     size_t size_of_header = sizeof(num_data_) + sizeof(num_features_) + sizeof(num_total_features_)
       + sizeof(int) * num_total_features_ + sizeof(label_idx_) + sizeof(num_groups_) + sizeof(sparse_threshold_)
       + 3 * sizeof(int) * num_features_ + sizeof(uint64_t) * (num_groups_ + 1) + 2 * sizeof(int) * num_groups_ + sizeof(int8_t) * num_features_
-      + sizeof(double) * num_features_ + sizeof(int) * 3 + sizeof(bool) * 2;
+      + sizeof(double) * num_features_ + sizeof(int32_t) * num_total_features_ + sizeof(int) * 3 + sizeof(bool) * 2;
     // size of feature names
     for (int i = 0; i < num_total_features_; ++i) {
       size_of_header += feature_names_[i].size() + sizeof(int);
@@ -682,6 +691,13 @@ void Dataset::SaveBinaryFile(const char* bin_filename) {
     if (ArrayArgs<double>::CheckAll(feature_penalty_, 1.0)) {
       feature_penalty_.clear();
     }
+    if (max_bin_by_feature_.empty()) {
+      ArrayArgs<int32_t>::Assign(&max_bin_by_feature_, -1, num_total_features_);
+    }
+    writer->Write(max_bin_by_feature_.data(), sizeof(int32_t) * num_total_features_);
+    if (ArrayArgs<int32_t>::CheckAll(max_bin_by_feature_, -1)) {
+      max_bin_by_feature_.clear();
+    }
     // write feature names
     for (int i = 0; i < num_total_features_; ++i) {
       int str_len = static_cast<int>(feature_names_[i].size());
@@ -730,6 +746,10 @@ void Dataset::DumpTextFile(const char* text_filename) {
   for (auto i : feature_penalty_) {
     fprintf(file, "%lf, ", i);
   }
+  fprintf(file, "\nmax_bin_by_feature: ");
+  for (auto i : max_bin_by_feature_) {
+    fprintf(file, "%d, ", i);
+  }
   fprintf(file, "\n");
   for (auto n : feature_names_) {
     fprintf(file, "%s, ", n.c_str());

src/io/dataset_loader.cpp

@@ -427,6 +427,23 @@ Dataset* DatasetLoader::LoadFromBinFile(const char* data_filename, const char* b
     dataset->feature_penalty_.clear();
   }
 
+  if (!config_.max_bin_by_feature.empty()) {
+    CHECK(static_cast<size_t>(dataset->num_total_features_) == config_.max_bin_by_feature.size());
+    CHECK(*(std::min_element(config_.max_bin_by_feature.begin(), config_.max_bin_by_feature.end())) > 1);
+    dataset->max_bin_by_feature_.resize(dataset->num_total_features_);
+    dataset->max_bin_by_feature_.assign(config_.max_bin_by_feature.begin(), config_.max_bin_by_feature.end());
+  } else {
+    const int32_t* tmp_ptr_max_bin_by_feature = reinterpret_cast<const int32_t*>(mem_ptr);
+    dataset->max_bin_by_feature_.clear();
+    for (int i = 0; i < dataset->num_total_features_; ++i) {
+      dataset->max_bin_by_feature_.push_back(tmp_ptr_max_bin_by_feature[i]);
+    }
+  }
+  mem_ptr += sizeof(int32_t) * (dataset->num_total_features_);
+  if (ArrayArgs<int32_t>::CheckAll(dataset->max_bin_by_feature_, -1)) {
+    dataset->max_bin_by_feature_.clear();
+  }
+
   // get feature names
   dataset->feature_names_.clear();
   // write feature names
@@ -544,7 +561,10 @@ Dataset* DatasetLoader::CostructFromSampleData(double** sample_values,
       feature_names_.push_back(str_buf.str());
     }
   }
-
+  if (!config_.max_bin_by_feature.empty()) {
+    CHECK(static_cast<size_t>(num_col) == config_.max_bin_by_feature.size());
+    CHECK(*(std::min_element(config_.max_bin_by_feature.begin(), config_.max_bin_by_feature.end())) > 1);
+  }
   const data_size_t filter_cnt = static_cast<data_size_t>(
     static_cast<double>(config_.min_data_in_leaf * total_sample_size) / num_data);
   if (Network::num_machines() == 1) {
@@ -562,8 +582,16 @@ Dataset* DatasetLoader::CostructFromSampleData(double** sample_values,
         bin_type = BinType::CategoricalBin;
       }
       bin_mappers[i].reset(new BinMapper());
-      bin_mappers[i]->FindBin(sample_values[i], num_per_col[i], total_sample_size,
-                              config_.max_bin, config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      if (config_.max_bin_by_feature.empty()) {
+        bin_mappers[i]->FindBin(sample_values[i], num_per_col[i], total_sample_size,
+                                config_.max_bin, config_.min_data_in_bin, filter_cnt, 
+                                bin_type, config_.use_missing, config_.zero_as_missing);
+      } else {
+        bin_mappers[i]->FindBin(sample_values[i], num_per_col[i], total_sample_size,
+                                config_.max_bin_by_feature[i], config_.min_data_in_bin, 
+                                filter_cnt, bin_type, config_.use_missing, 
+          config_.zero_as_missing);
+      }
       OMP_LOOP_EX_END();
     }
     OMP_THROW_EX();
@@ -599,8 +627,16 @@ Dataset* DatasetLoader::CostructFromSampleData(double** sample_values,
         bin_type = BinType::CategoricalBin;
       }
       bin_mappers[i].reset(new BinMapper());
-      bin_mappers[i]->FindBin(sample_values[start[rank] + i], num_per_col[start[rank] + i], total_sample_size,
-                              config_.max_bin, config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      if (config_.max_bin_by_feature.empty()) {
+        bin_mappers[i]->FindBin(sample_values[start[rank] + i], num_per_col[start[rank] + i], 
+                                total_sample_size, config_.max_bin, config_.min_data_in_bin, 
+                                filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      } else {
+        bin_mappers[i]->FindBin(sample_values[start[rank] + i], num_per_col[start[rank] + i], 
+                                total_sample_size, config_.max_bin_by_feature[start[rank] + i], 
+                                config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, 
+                                config_.zero_as_missing);
+      }
       OMP_LOOP_EX_END();
     }
     OMP_THROW_EX();
@@ -831,6 +867,11 @@ void DatasetLoader::ConstructBinMappersFromTextData(int rank, int num_machines,
     dataset->num_total_features_ = static_cast<int>(feature_names_.size());
   }
 
+  if (!config_.max_bin_by_feature.empty()) {
+    CHECK(static_cast<size_t>(dataset->num_total_features_) == config_.max_bin_by_feature.size());
+    CHECK(*(std::min_element(config_.max_bin_by_feature.begin(), config_.max_bin_by_feature.end())) > 1);
+  }
+
   // check the range of label_idx, weight_idx and group_idx
   CHECK(label_idx_ >= 0 && label_idx_ <= dataset->num_total_features_);
   CHECK(weight_idx_ < 0 || weight_idx_ < dataset->num_total_features_);
@@ -865,8 +906,16 @@ void DatasetLoader::ConstructBinMappersFromTextData(int rank, int num_machines,
         bin_type = BinType::CategoricalBin;
       }
       bin_mappers[i].reset(new BinMapper());
-      bin_mappers[i]->FindBin(sample_values[i].data(), static_cast<int>(sample_values[i].size()),
-                              sample_data.size(), config_.max_bin, config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      if (config_.max_bin_by_feature.empty()) {
+        bin_mappers[i]->FindBin(sample_values[i].data(), static_cast<int>(sample_values[i].size()),
+                                sample_data.size(), config_.max_bin, config_.min_data_in_bin, 
+                                filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      } else {
+        bin_mappers[i]->FindBin(sample_values[i].data(), static_cast<int>(sample_values[i].size()),
+                                sample_data.size(), config_.max_bin_by_feature[i], 
+                                config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, 
+                                config_.zero_as_missing);
+      }
       OMP_LOOP_EX_END();
     }
     OMP_THROW_EX();
@@ -902,8 +951,18 @@ void DatasetLoader::ConstructBinMappersFromTextData(int rank, int num_machines,
         bin_type = BinType::CategoricalBin;
       }
       bin_mappers[i].reset(new BinMapper());
-      bin_mappers[i]->FindBin(sample_values[start[rank] + i].data(), static_cast<int>(sample_values[start[rank] + i].size()),
-                              sample_data.size(), config_.max_bin, config_.min_data_in_bin, filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      if (config_.max_bin_by_feature.empty()) {
+        bin_mappers[i]->FindBin(sample_values[start[rank] + i].data(), 
+                                static_cast<int>(sample_values[start[rank] + i].size()),
+                                sample_data.size(), config_.max_bin, config_.min_data_in_bin, 
+                                filter_cnt, bin_type, config_.use_missing, config_.zero_as_missing);
+      } else {
+        bin_mappers[i]->FindBin(sample_values[start[rank] + i].data(), 
+                                static_cast<int>(sample_values[start[rank] + i].size()),
+                                sample_data.size(), config_.max_bin_by_feature[i], 
+                                config_.min_data_in_bin, filter_cnt, bin_type, 
+          config_.use_missing, config_.zero_as_missing);
+      }
       OMP_LOOP_EX_END();
     }
     OMP_THROW_EX();

tests/python_package_test/test_engine.py

@@ -813,6 +813,29 @@ def is_correctly_constrained(learner):
         constrained_model = lgb.train(params, trainset)
         self.assertTrue(is_correctly_constrained(constrained_model))
 
+    def test_max_bin_by_feature(self):
+        col1 = np.arange(0, 100)[:, np.newaxis]
+        col2 = np.zeros((100, 1))
+        col2[20:] = 1
+        X = np.concatenate([col1, col2], axis=1)
+        y = np.arange(0, 100)
+        params = {
+            'objective': 'regression_l2',
+            'verbose': -1,
+            'num_leaves': 100,
+            'min_data_in_leaf': 1,
+            'min_sum_hessian_in_leaf': 0,
+            'min_data_in_bin': 1,
+            'max_bin_by_feature': [100, 2]
+        }
+        lgb_data = lgb.Dataset(X, label=y)
+        est = lgb.train(params, lgb_data, num_boost_round=1)
+        self.assertEqual(len(np.unique(est.predict(X))), 100)
+        params['max_bin_by_feature'] = [2, 100]
+        lgb_data = lgb.Dataset(X, label=y)
+        est = lgb.train(params, lgb_data, num_boost_round=1)
+        self.assertEqual(len(np.unique(est.predict(X))), 3)
+
     def test_refit(self):
         X, y = load_breast_cancer(True)
         X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)