Summary
Scalding provides a number of ways to implement and execute machine learning algorithms. As presented, we can manipulate pipes, use the Matrix API or algebird, and interoperate with existing libraries such as Mahout.
The majority of ML jobs originate as Big Data ETL jobs that reduce to a smaller data space. The final result usually needs some form of post-processing, and it is then stored in an external source. Scalding provides great interoperability with external systems, and it is thus one of the most suitable technologies to solve such problems.
