Open
Description
We could consider allowing callbacks that enables custom, user-defined iteration stopping.
For example, there are cases where users may want to compare unfolded distributions to data distributions and stop unfolding iterations when some condition is met (e.g. a bias-variance tradeoff). I can imagine there are many possible stopping conditions, perhaps even different stopping conditions are appropriate for different analyses. It would be nice to support these situations.
Here's one potential way we could support user-defined stopping conditions:
from pyunfold.callbacks import Callback
# Create custom callback that will stop unfolding iterations when some condition is met
class CustomStoppingCondition(Callback):
def __init__(self):
super(Callback, self).__init__()
pass
def on_iteration_end(self, iteration, status):
# Define stopping condition logic here
stopping_condition_met = ...
if stopping_condition_met:
# stop iterating
else:
# keep iterating
iteration_stopping = CustomStoppingCondition()
unfolded_result = iterative_unfold(data=data,
data_err=data_err,
response=response,
response_err=response_err,
efficiencies=efficiencies,
efficiencies_err=efficiencies_err,
callbacks=iteration_stopping)Personally, I think this would fit nicely into the existing callback interface. I'm curious to see what other think about this.