diff --git a/examples/mnist/lenet-stn.jl b/examples/mnist/lenet-stn.jl
index 23ca9de3f..60f2def68 100644
--- a/examples/mnist/lenet-stn.jl
+++ b/examples/mnist/lenet-stn.jl
@@ -12,16 +12,16 @@ data = mx.Variable(:data)
 # the localisation network in lenet-stn
 # it will increase acc about more than 1%, when num-epoch >=15
 # The localization net just takes the data as input and must output a vector in R^n
-loc_net = @mx.chain mx.Convolution(data, num_filter=10, kernel=(5, 5), stride=(2,2)) =>
+loc_net = @mx.chain mx.Convolution(data, num_filter=30, kernel=(5, 5), stride=(2, 2)) =>
                     mx.Activation(act_type=:relu) =>
-                    mx.Pooling( kernel=(2, 2), stride=(2, 2), pool_type=:max) =>
-                    mx.Convolution( num_filter=10, kernel=(3, 3), stride=(2,2), pad=(1, 1)) =>
+                    mx.Pooling(kernel=(2, 2), stride=(2, 2), pool_type=:max) =>
+                    mx.Convolution(num_filter=60, kernel=(3, 3), stride=(2, 2), pad=(1, 1)) =>
                     mx.Activation(act_type=:relu) =>
-                    mx.Pooling( global_pool=true, kernel=(2, 2), pool_type=:avg) =>
+                    mx.Pooling(global_pool=true, kernel=(2, 2), pool_type=:avg) =>
                     mx.Flatten() =>
                     mx.FullyConnected(num_hidden=6, name=:stn_loc)
 
-data=mx.SpatialTransformer(data,loc_net, target_shape = (28,28), transform_type="affine", sampler_type="bilinear")
+data = mx.SpatialTransformer(data, loc_net, target_shape=(28, 28), transform_type="affine", sampler_type="bilinear")
 
 # first conv
 conv1 = @mx.chain mx.Convolution(data, kernel=(5,5), num_filter=20)  =>
@@ -57,8 +57,13 @@ train_provider, eval_provider = get_mnist_providers(batch_size; flat=false)
 model = mx.FeedForward(lenet, context=mx.cpu())
 
 # optimizer
-optimizer = mx.ADAM(lr=0.01, weight_decay=0.00001)
+optimizer = mx.SGD(lr=0.1, momentum=.9)
 
 # fit parameters
-initializer=mx.XavierInitializer(distribution = mx.xv_uniform, regularization = mx.xv_avg, magnitude = 1)
-mx.fit(model, optimizer, train_provider, n_epoch=20, eval_data=eval_provider,initializer=initializer)
+initializer = mx.XavierInitializer(distribution=mx.xv_normal,
+                                   regularization=mx.xv_in,
+                                   magnitude=2)
+mx.fit(model, optimizer, train_provider,
+       n_epoch=20,
+       eval_data=eval_provider,
+       initializer=initializer)