You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It will often be useful or necessary to treat topological features (persistence barcodes) of each homological degree separately, e.g. to learn the scale sequences over which to vectorize or to apply different vectorizations. This could be facilitated in at least 3 ways:
Add a degree parameter to step_vpd_*(), which will filter the input PDs before vectorization.
Write a new step (step_phom_degree()?) to take a PD column as input and output separate columns for specified degrees.
Add a logical parameter (by_degree?) to step_phom_*() to control whether the step outputs a single column of PDs or a PD column for each degree.
I believe (3) is most natural; but either (1) or (2) will be needed in case some vectorizations receive the full PDs while others receive only features of a specified degree. So i plan to implement both (1) and (3), to avoid having to create a whole new step.
Feedback is welcome!
The text was updated successfully, but these errors were encountered:
It will often be useful or necessary to treat topological features (persistence barcodes) of each homological degree separately, e.g. to learn the scale sequences over which to vectorize or to apply different vectorizations. This could be facilitated in at least 3 ways:
degree
parameter tostep_vpd_*()
, which will filter the input PDs before vectorization.step_phom_degree()
?) to take a PD column as input and output separate columns for specified degrees.by_degree
?) tostep_phom_*()
to control whether the step outputs a single column of PDs or a PD column for each degree.I believe (3) is most natural; but either (1) or (2) will be needed in case some vectorizations receive the full PDs while others receive only features of a specified degree. So i plan to implement both (1) and (3), to avoid having to create a whole new step.
Feedback is welcome!
The text was updated successfully, but these errors were encountered: