That tuning also depends on whether your new model has new data (re: “online learning”), in that it may change the scaling of your posterior.
I don’t know how much the tuning depends on the (new) data, but my intuition tells me “maybe a lot”.
1 Like