Continuous Learning with ScaleOut Digital Twins

Depending on your algorithm type or work flow, you have several options to update your machine learning algorithms using ScaleOut Digital Twins. Retraining an algorithm requires new training data. The digital twins can use an API to report anomalies in case of identifiable issues. These new data points are stored and available for download. This new data can be added to the original dataset to train the algorithm from scratch using an extended dataset, or if the algorithm supports incremental retraining, the new dataset can be fed to the original trained algorithm for refining.

In the next sections, we will describe how to:

• gather and download new training data (and retrain your machine learning algorithm outside of ScaleOut Digital Twins)

• gather new training data and incrementally retrain your machine learning algorithm without the need to redeploy

• manually update a machine learning algorithm for a deployed digital twin model