Data Frames

Data Frames are used to store data during execution of an ML pipeline. They are similar to a SQL table in that they have a schema for storing the data types of every column and they have rows for storing the actual values.

Spark, Scikit-learn, and MLeap all have their own version of a data frame. Tensorflow uses a graph of inputs and outputs to execute transformations, which is very easy to inteface with a data frame structure.

Spark Data Frames

Spark's data frames are optimized for distributed computation, making them excellent for processing large datasets. They are very heavy-weight, as they need to handle network failure scenarios, compilation of execution plans, redundancy, and many other requirements in a distributed context. Spark data frames offer a lot of functionality outside of ML pipelines, such as joining large datasets, mapping, reducing, SQL queries, etc.

Scikit-learn Data Frames

Scikit-learn data frames are provided by Pandas and NumPy. These are lightweight data structures, and offer quite a bit of the same functionality as Spark data frames, minus the distributed nature of Spark's data frames.

MLeap Data Frames: Leap Frames

Leap frames are very lightweight data structures and are meant to support very basic operations and ML transformations. Because of their simplicity, they are highly-optimized for use as a realtime prediction engine or small-batch predictions. Leap frames can be abstracted over Spark data frames, and so they do not lose their ability to act as an efficient batch-mode data store as well.

Example Leap Frame

Here is an example leap frame in JSON, it comes from our AirBnB demo:

{
  "schema": {
    "fields": [{
      "name": "state",
      "type": "string"
    }, {
      "name": "bathrooms",
      "type": "double"
    }, {
      "name": "square_feet",
      "type": "double"
    }, {
      "name": "bedrooms",
      "type": "double"
    }, {
      "name": "review_scores_rating",
      "type": "double"
    }, {
      "name": "room_type",
      "type": "string"
    }, {
      "name": "cancellation_policy",
      "type": "string"
    }]
  },
  "rows": [["NY", 2.0, 1250.0, 3.0, 50.0, "Entire home/apt", "strict"]]
}

Tensorflow

Tensorflow does not have data frames like Spark, Scikit-learn and MLeap. Instead, Tensorflow relies on input nodes and output nodes, connected by a graph of transfomation operations. This paradigm is actually neatly compatible with data frames, as certain columns can be used to provide data for the input nodes, while output nodes can be placed in new columns of a data frame. Leap frames are specifically designed to be compatible with Tensorflow graphs, Spark data frames, and to a certain extent, Scikit-learn data frames.