Online learning for time-series
There are two main scenarios of learning – online learning and offline learning. Online learning means that you are fitting your model incrementally as the data flows in (streaming data). On the other hand, offline learning, the more commonly known approach, implies that you have a static dataset that you know from the start, and the parameters of your machine learning algorithm are adjusted to the whole dataset at once (often loading the whole dataset into memory or in batches).
There are three major use cases for online learning:
Typically, in online learning settings, you have more data, and it is appropriate for big data. Online learning can be applied to large datasets, where it would be computationally infeasible to train over the entire dataset.
Another use case for online learning is where the inference and fitting are performed...