Fix Tutorial 201B for convergence issue.

Documentation/current_iteration.md

@@ -18,3 +18,6 @@ C.debugging.stop_profiler()
 ## CPU inference performance improvements using MKL
 - Accelerates some common tensor ops in Intel CPU inference for float32, especially for fully connected networks
 - Can be turned on/off by cntk.cntk_py.enable_cpueval_optimization()/cntk.cntk_py.disable_cpueval_optimization()
+
+## Bug fixes
+- Fixed convergence issue in Tutorial 201B

Tests/EndToEndTests/CNTKv2Python/Tutorials/CNTK_201B_CIFAR-10_ImageHandsOn_test.py

@@ -17,3 +17,18 @@ def test_cntk_201B_cifar_10_imagehandson_noErrors(nb):
     errors = [output for cell in nb.cells if 'outputs' in cell
               for output in cell['outputs'] if output.output_type == "error"]
     assert errors == []
+
+    metrics = []
+    for cell in nb.cells:
+        try:
+           if cell.cell_type == 'code':
+               m = re.search('Final Results: .* errs = (?P<metric>\d+\.\d+)%', cell.outputs[0]['text'])
+               if m:
+                   metrics.append(float(m.group('metric')))
+                   break
+        except IndexError:
+           pass
+        except KeyError:
+           pass
+    # TODO tighten tolerances
+    assert np.allclose([43.3], metrics, atol=0.5)

Tutorials/CNTK_201B_CIFAR-10_ImageHandsOn.ipynb

@@ -521,8 +521,8 @@
     "\n",
     "    # Set training parameters\n",
     "    lr_per_minibatch       = C.learning_parameter_schedule([0.01]*10 + [0.003]*10 + [0.001], \n",
-    "                                                       epoch_size)\n",
-    "    momentums              = C.momentum_schedule(0.9, minibatch_size)\n",
+    "                                                       epoch_size = epoch_size)\n",
+    "    momentums              = C.momentum_schedule(0.9, minibatch_size = minibatch_size)\n",
     "    l2_reg_weight          = 0.001\n",
     "    \n",
     "    # trainer object\n",