=================================================================================

Data from a tf.data.Dataset can be taken to refactor linear regression, and then implement stochastic gradient descent with it. In this case, the dataset will be synthetic and read by the tf.data API directly from memory, and then tf.data API is used to load a dataset when the dataset resides on disk. The steps for this application are:
            iv.a) Use tf.data to read data from memory.
            iv.b) Use tf.data in a training loop.
            iv.c) Use tf.data to read data from disk.
            iv.d) Write production input pipelines with features engineering (batching, shuffling, etc.)

============================================