diff --git a/wgan2-AC-lasagne.ipynb b/wgan2-AC-lasagne.ipynb
index 6e35596..a138e55 100644
--- a/wgan2-AC-lasagne.ipynb
+++ b/wgan2-AC-lasagne.ipynb
@@ -324,16 +324,16 @@
     "loss_D_real = output_D.mean()\n",
     "loss_D_fake = output_D_fake.mean()\n",
     "loss_D = loss_D_fake - loss_D_real\n",
-    "loss_D_gp = loss_D + 10 * grad_penalty + loss_C_real\n",
+    "loss_D_gp = loss_D + 10 * grad_penalty + 0.01*loss_C_real\n",
     "loss_G = -loss_D_fake + loss_C_fake\n",
     "\n",
     "\n",
     "params_netD = lasagne.layers.get_all_params(netD, trainable=True) \n",
     "params_netC = lasagne.layers.get_all_params(netC, trainable=True) \n",
     "params_netG = lasagne.layers.get_all_params(netG, trainable=True)\n",
-    "optimize_G = lasagne.updates.adam(loss_G, params_netG, learning_rate=lrG, beta1=0.2)\n",
+    "optimize_G = lasagne.updates.adam(loss_G, params_netG, learning_rate=lrG, beta1=0.)\n",
     "optimize_D = lasagne.updates.adam(loss_D_gp, params_netD+params_netC, \n",
-    "                                  learning_rate=lrD, beta1=0.2)\n",
+    "                                  learning_rate=lrD, beta1=0.)\n",
     "train_G_fn =  theano.function([input_var_G], [loss_G], updates=optimize_G)\n",
     "train_D_fn = theano.function([input_var_D, input_var_G, input_var_Y, ϵ], \n",
     "                              [loss_D_gp, loss_D, loss_D_real, loss_D_fake, loss_C_fake], \n",