signal processing - How do I evaluate a model that applies a time delay to the predictions? -
so i'm working on data science project uses bayesian model predictions , have used median filter smoothen/filter output. however, have noticed effect of filter output receive ground truth data time-shifted forward half window length of filter. purpose of experiment, not mind having data 7 seconds late (my filter's window length 15) long more accurate.
i not sure if acceptable correlate output shifted 7(.5) seconds ground truth data calculate accuracy or if should report accuracy result without me meddling output vectors.
Comments
Post a Comment