fix the bug in bin with small values (#2342)

* fix the bug in bin with small values

* Update bin.cpp

* Update test_engine.py

src/io/bin.cpp

@@ -181,6 +181,9 @@ namespace LightGBM {
       int left_max_bin = static_cast<int>(static_cast<double>(left_cnt_data) / (total_sample_cnt - cnt_zero) * (max_bin - 1));
       left_max_bin = std::max(1, left_max_bin);
       bin_upper_bound = GreedyFindBin(distinct_values, counts, left_cnt, left_max_bin, left_cnt_data, min_data_in_bin);
+      if (bin_upper_bound.size() > 0) {
+        bin_upper_bound.back() = -kZeroThreshold;
+      }
     }
 
     int right_start = -1;
@@ -191,32 +194,16 @@ namespace LightGBM {
       }
     }
 
-    if (bin_upper_bound.size() == 0) {
-      if (max_bin > 2) {
-        // create zero bin
-        bin_upper_bound.push_back(-kZeroThreshold);
-        bin_upper_bound.push_back(kZeroThreshold);
-      }
-      else if (max_bin > 1) {
-        bin_upper_bound.push_back(kZeroThreshold);
-      }
-    } else {
-      bin_upper_bound.back() = -kZeroThreshold;
-      if (max_bin > 2) {
-        // create zero bin
-        bin_upper_bound.push_back(kZeroThreshold);
-      }
-    }
-
-    int right_max_bin = max_bin - static_cast<int>(bin_upper_bound.size());
-    if ((right_start >= 0) && (right_max_bin > 0)) {
+    int right_max_bin = max_bin - 1 - static_cast<int>(bin_upper_bound.size());
+    if (right_start >= 0 && right_max_bin > 0) {
       auto right_bounds = GreedyFindBin(distinct_values + right_start, counts + right_start,
         num_distinct_values - right_start, right_max_bin, right_cnt_data, min_data_in_bin);
+      bin_upper_bound.push_back(kZeroThreshold);
       bin_upper_bound.insert(bin_upper_bound.end(), right_bounds.begin(), right_bounds.end());
     } else {
       bin_upper_bound.push_back(std::numeric_limits<double>::infinity());
     }
-    CHECK(bin_upper_bound.size() <= max_bin);
+    CHECK(bin_upper_bound.size() <= static_cast<size_t>(max_bin));
     return bin_upper_bound;
   }
 

tests/python_package_test/test_engine.py

@@ -921,7 +921,7 @@ class TestEngine(unittest.TestCase):
         }
         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))), 99)
+        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)