search

pgmpy / pgmpy_notebook

Short Tutorial to Probabilistic Graphical Models(PGM) and pgmpy
http://pgmpy.org/
MIT License
368 stars 213 forks source link

Updating fitted model based on new upcoming data #47

Open LorenzoMonni opened 3 years ago

LorenzoMonni commented 3 years ago

This is not related to a bug in the code, but instead to an idea of a functionality of models. Say we have a model that was already created from a dataset with an estimator, is there an automatic way to update this model based on new data samples instead of appending the new samples to the old dataset and having to run the fitting again with the full dataset from the beginning?

ankurankan commented 3 years ago

@LorenzoMonni There is no direct way to update the model parameters with new data yet. I will try to have this feature in next to next release. But in the meanwhile, you can simulate the updating behavior using BayesianEstimators and Dirichlet prior. I have shown an example here: https://github.com/pgmpy/pgmpy/issues/1350#issuecomment-726724566%EF%BC%8Cmy

LorenzoMonni commented 3 years ago

Ok, thank you so much!