diff --git a/src/operator/batch_norm-inl.h b/src/operator/batch_norm-inl.h
index 01a478429ce9..89f5327c9722 100755
--- a/src/operator/batch_norm-inl.h
+++ b/src/operator/batch_norm-inl.h
@@ -31,6 +31,7 @@ struct BatchNormParam : public dmlc::Parameter<BatchNormParam> {
   float eps;
   float momentum;
   bool fix_gamma;
+  bool fix_linear_trans;
   bool use_global_stats;
   bool output_mean_var;
   DMLC_DECLARE_PARAMETER(BatchNormParam) {
@@ -40,6 +41,8 @@ struct BatchNormParam : public dmlc::Parameter<BatchNormParam> {
     .describe("Momentum for moving average");
     DMLC_DECLARE_FIELD(fix_gamma).set_default(true)
     .describe("Fix gamma while training");
+    DMLC_DECLARE_FIELD(fix_linear_trans).set_default(false)
+    .describe("Fix linear transformation while training");
     DMLC_DECLARE_FIELD(use_global_stats).set_default(false)
     .describe("Whether use global moving statistics instead of local batch-norm. "
               "This will force change batch-norm into a scale shift operator.");
@@ -180,7 +183,7 @@ class BatchNormOp : public Operator {
       tmp *= gvar;
       gmean += tmp;
       // assign
-      if (!param_.fix_gamma) {
+      if (!param_.fix_gamma || !param_.fix_linear_trans) {
         Assign(gslope, req[batchnorm::kGamma],
                sumall_except_dim<1>(
                    grad * (data - broadcast<1>(mean, data.shape_)) /
@@ -197,7 +200,7 @@ class BatchNormOp : public Operator {
       Assign(gbias, req[batchnorm::kBeta], sumall_except_dim<1>(grad));
     } else {
       // use global statistics with freeze moving mean and var.
-      if (!param_.fix_gamma) {
+      if (!param_.fix_gamma || !param_.fix_linear_trans) {
         Assign(gslope, req[batchnorm::kGamma],
                sumall_except_dim<1>(
                    grad * (data - broadcast<1>(moving_mean, data.shape_)) /