Being pretty new to ML.Net and working with TimeSeriesPrediction models and engine, I don't quite get and find enough documentation about what exactly TimeSeriesPredictionEngine<TSrc,TDst>.CheckPoint method does.
What I understand is that when we have a new feed in the time series we should checkpoint the model for next prediction, but can we keep doing this forever? Do we eventually need to retrain the model after a while or using the Checkpoint methods does the same thing as retraining the model with all the history?
What are the best practices in your experience?
Asked
Active
Viewed 121 times
3
-
I don't have a proper answer but the docs seem to have some useful samples: https://learn.microsoft.com/en-us/dotnet/api/microsoft.ml.transforms.timeseries.timeseriespredictionengine-2.checkpoint?view=ml-dotnet – keithl8041 Mar 11 '22 at 19:56
-
It has some good examples but it doesn't answer this question. – CageE Mar 12 '22 at 15:34