Tickets/dm 9828 #72
Tickets/dm 9828 #72
Conversation
yalsayyad
force-pushed
the
tickets/DM-9828
branch
from
0345db5
to
7a4e1b0
Compare
| @param[in] algorithm name of interpolation algorithm; if None use self.config.algorithm | ||
|
|
||
| @return fit background as an lsst.afw.math.Background | ||
|
|
||
| @throw RuntimeError if lsst.afw.math.makeBackground returns None, | ||
| which is apparently one way it indicates failure | ||
| """ | ||
|
|
||
| binSizeX = self.config.binSizeX if self.config.binSizeX else self.config.binSize | ||
| binSizeY = self.config.binSizeY if self.config.binSizeY else self.config.binSize |
There was a problem hiding this comment.
Add explicit == 0? Makes it clearer and corresponds more exactly with your config docs.
| elif self.config.undersampleStyle == "REDUCE_INTERP_ORDER": | ||
| if order < 1: | ||
| raise ValueError("Cannot reduce approxOrder below 0. " + | ||
| "Try using undersampleStyle = \"INCREASE_NXNYSAMPLE\" instead?") | ||
| order = min(nx, ny) - 1 | ||
| self.log.warn("Reducing approxOrder to %d" % order) | ||
| elif self.config.undersampleStyle == "INCREASE_NXNYSAMPLE": | ||
| # Reduce bin size to the largest acceptable square bins |
There was a problem hiding this comment.
Do they have to be square? The order is (unfortunately) identical for both dimensions, but I don't think there's a reason the bin size has to be identical.
There was a problem hiding this comment.
They don't. A goal with this ticket was not to change behavior of existing configs. As I mentioned on the ticket comments, its not obvious what behavior is more intuitive and expected here: to reduce the bin size only in the dimension with too few data points or reduce the bin size in both dimensions. Given that everyone up until this point has specified square bins, I thought it would be least surprising if the bins were reduced in both dimensions as they were before this ticket.
yalsayyad
force-pushed
the
tickets/DM-9828
branch
from
7a4e1b0
to
0740463
Compare
Add config parameters binSizeX and binSizeY which control the bin size in the x and y dimensions individually. If either is 0, the default, the general binSize parameter is used for the bin size in that dimension. Applies to both polynomial and spline background models.
of order and config.approxOrderX in the undersampling check in SubtractBackgroundTask. This change does not alter behavior because the order is assigned config.approxOrderX.
Chebyshev polynomials are the default background model, but when splines are requested, the default interpolation style should be the more stable and robust AKIMA_SPLINE, to be consistent with other background subtraction classes and documentation.
yalsayyad
force-pushed
the
tickets/DM-9828
branch
from
0740463
to
c795208
Compare
No description provided.