Kipp Cannon authored 6 years ago and Kipp Cannon committed 6 years ago

- the count is meaningless, it is only used to set the scale of the density estimation kernel, but for that reason it should not change as the result of addition operations
- in the current pipeline care is taken to never add more than a single copy of the numerator into any ranking statistic in order to control the total count
- this patch causes .__iadd__() to preserve the total count, allowing the analysis pipeline plumbing to be simplified since there is no reason to fear adding multiple copies of the numerator together

- when initializing an SNR PDF cache with new entries, a math domain error
  can occur when an instrument combination is impossible.  this patch fixes
  it
- it's very odd that this code worked in the past.  Python must be changing
  its behaviour.  commit 29a7482f contains
  both the original version of this code and the data file it successfully
  generated at the time ... so ... it must have worked.  I don't know what
  else to say, but I'm not crazy.

- switch numerator and denominator PDFs to a rate.BinnedLnPDF + snglcoinc.LnLRDensity() -based implementation
- add trigger rate tracking feature to denominator, taking advantage of new high-speed coincidence rate machinery in snglcoinc
- add signal rate tracking feature to numerator, taking advantage of pre-computed P(instruments) factors in SNR PDF look-up table
- switch ranking statistic distribution to a rate.BinnedLnPDF -based implementation
- remove search_summary tables, attach all livetime tracking to the trigger rate data in the denominator object

this is a work in progress.  some notable features that are still missing:

- --min-log-L cut based candidate rate reduction feature in gstlal_inspiral is disabled
- online mode for gstlal_inspiral is non-functional

Kipp Cannon authored 7 years ago and Kipp Cannon committed 7 years ago

- subclass of rate.BinnedLnPDF to implement 1-parameter family of 1-D PDFs P(\chi^2 | snr, signal)

- introduce internal cacheentry class to hold entry information
- remove limit on cache size
- when initializing, move loop over instruments into .add_to_cache()
- add min_instruments parameter to the cache key
- in P_instruments_given_signal(), handle off instruments better
- record P(instruments | signal) in the SNRPDF cache
- enables generalization of ranking statistic away from fixed P(instruments | signal) marginalized over horizon distance history

- and change message from "... SNR PDF" to "... candidates"

- instead of a default keyword argument of .quantize_horizon_distances(), store it as an instance attribute
- store the value in the XML PDF cache, and retrieve on load
- update pre-computed cache in share/ to new format (= manually add min_ratio Param to XML tree)

- some of this stuff is almost certainly broken, but I'm not trying to fix it I'm just removing the word "pylal" from our codebase.

- glue.ligolw.lsctables.instrument_set_from_ifos() --> instrumentsproperty.get()
- glue.ligolw.lsctables.ifos_from_instrument_set() --> instrumentsproperty.set()

- remove comparefunc and max_dt arguments from ligolw_thinca() call
- remove local custom EventList implementation, ligolw_thinca no implements this logic upstream

- clean up lnP_dt_dphi_uniform_H1L1():  don't use mixed integer/float expressions, reduce number of arithmetic operations
- delete coinc_window() and port random_params() to snglcoinc.CoincSynthesizer:  works for any number of isntruments
- add/update FIXMEs related to sampler density normalization

- NearestLeafTree()
- instruments_rate_given_noise()
- P_instruments_given_signal()

- pylal.inject is dead code

- Python standard library now provides drop-in replacement

distribution

defined to be the difference between the coinc inspiral and single inspiral end
times