What is online machine learning?
In machine learning, the most common way to train a model is to do a single training pass. The general steps in this approach are as follows:
- Data preparation.
- Create a train-test split.
- Do model benchmarking and hyperparameter tuning.
- Select the best model.
- Move this model to production.
This approach is called offline learning.
With streaming data, you can often use this type of model very well. You can build the model once, deploy it, and use it for predicting your input stream. You can probably track the performance metrics of your model, and when the performance starts to change, you can do an update or retraining of your model, deploy the new version, and let it set in the production environment as long as it works.
Online machine learning is a branch of machine learning that contains models that work very differently. They do not learn a full dataset at once, but rather, update the learned model (the decision...
