PM + team feedback

Author: nibaccam

articles/machine-learning/concept-distributed-training.md

@@ -19,13 +19,13 @@ In distributed training the workload to train a model is split up and shared amo
 
 ## Deep learning and distributed training 
 
-There are two main types of distributed training: [data parallelism](#data-parallelism) and [model parallelism](#model-parallelism). For distributed training on deep learning models, the [Azure Machine Learning SDK in Python](https://docs.microsoft.com/python/api/overview/azure/ml/intro?view=azure-ml-py) supports integrations with popular frameworks, PyTorch and TensorFlow. Both frameworks employ data parallelism for distributed training, and leverage [horovod](https://horovod.readthedocs.io/en/latest/summary_include.html) for optimizing compute speeds. 
+There are two main types of distributed training: [data parallelism](#data-parallelism) and [model parallelism](#model-parallelism). For distributed training on deep learning models, the [Azure Machine Learning SDK in Python](https://docs.microsoft.com/python/api/overview/azure/ml/intro?view=azure-ml-py) supports integrations with popular frameworks, PyTorch and TensorFlow. Both frameworks employ data parallelism for distributed training, and can leverage [horovod](https://horovod.readthedocs.io/en/latest/summary_include.html) for optimizing compute speeds. 
 
 * [Distributed training with PyTorch](how-to-train-pytorch.md#distributed-training)
 
 * [Distributed training with TensorFlow](how-to-train-tensorflow.md#distributed-training)
 
-For training ML models that don't require distributed training, see [train models with Azure Machine Learning](concept-train-machine-learning-model.md#python-sdk) for the different ways to train models using the Python SDK.
+For ML models that don't require distributed training, see [train models with Azure Machine Learning](concept-train-machine-learning-model.md#python-sdk) for the different ways to train models using the Python SDK.
 
 ## Data parallelism
 
@@ -42,5 +42,6 @@ In model parallelism, worker nodes only need to synchronize the shared parameter
 ## Next steps
 
 * Learn how to [set up training environments](how-to-set-up-training-targets.md) with the Python SDK.
+* For a technical example, see the [reference architecture  scenario](https://docs.microsoft.com/azure/architecture/reference-architectures/ai/training-deep-learning).
 * [Train ML models with TensorFlow](how-to-train-tensorflow.md).
 * [Train ML models with PyTorch](how-to-train-pytorch.md).