diff --git a/src/rbm/state.py b/src/rbm/state.py
index 1dc557f..3058bc9 100644
--- a/src/rbm/state.py
+++ b/src/rbm/state.py
@@ -55,7 +55,7 @@ class RbmState:
 		np.savez(filename, whv=self.w_hv, bv=self.b_v, bh=self.b_h)
 		print(f"{filename} saved successfully!")
 
-	def init(self, mu: float = 0.5, std: float = 1.0):
+	def init(self, mu: float = 0.0, std: float = 1.0):
 		self.w_hv = uniform(self.w_hv.shape, mu, std)
 		self.b_v = uniform(self.b_v.shape, 0, 0)
 		self.b_h = uniform(self.b_h.shape, 0, 0)
diff --git a/src/rbm/train.py b/src/rbm/train.py
index 0573dab..0f53149 100644
--- a/src/rbm/train.py
+++ b/src/rbm/train.py
@@ -218,9 +218,9 @@ def train(entity: Entity, batch: Mat, status: Status):
 		if not keep_running:
 			break
 
+		# Reset gradient
+		entity.grad_zero()
 		for mini_batch in to_mini_batch(batch, mini_batch_size):
-			# Reset gradient
-			entity.grad_zero()
 			# Contrastive divergence learning: calculate gradients
 			dwhv, dbv, dbh = cd_func(entity, mini_batch)