Apache Spark 2.0 Preview: Machine Learning Model Persistence

This DataFrame-based API for MLlib provides functionality for saving and loading models that mimics the familiar Spark Data Source API. We will demonstrate saving and loading models in several languages using the popular MNIST dataset for handwritten digit recognition (LeCun et al., 1998; available from the LibSVM dataset page). This dataset contains handwritten digits 0–9, plus the ground truth labels.