Missing least-squares problems in constrained form
Here's a list of nonlinear SUR2
problems that currently don't have a feasibility problem equivalent:
- BDQRTIC
- BROWNAL
- BRYBND
- CHAINWOO
- CHNROSNB
- CHNRSNBM
- CUBE
- CURLY10
- CURLY20
- CURLY30
- DECONVU
- ERRINROS
- ERRINRSM
- EXTROSNB
- FREUROTH
- GENHUMPS
- GENROSE
- LIARWHD
- MANCINO
- MODBEALE
- MOREBV
- NONDIA
- NONMSQRT
- OSBORNEB
- OSCIGRAD
- OSCIPATH
- PENALTY1
- PENALTY2
- SBRYBND
- SCURLY10
- SCURLY20
- SCURLY30
- SENSORS
- SROSENBR
- SSBRYBND
- TQUARTIC
- VAREIGVL
- WATSON
- WOODS
Comments (8)
-
repo owner -
repo owner Additionally, BROWNALE is the equations version of BROWNAL, DECONVNE is that for DECONVU, FREURONE for FREUROTH, CHNRSBNE for CHNROSNB, MOREBVNE for MOREBV, OSBORNE2 for OSBORNEB, PENLT1NE for PENALTY1, PENLT2NE for PENALTY2, WATSONNE for WATSON, WOODSNE for WOODS, and CUBENE for CUBE.
Phew, the list of problems needing equation variants is now the more manageable
- BDQRTIC
- BRYBND
- CHAINWOO
- CHNRSNBM
- ERRINROS
- ERRINRSM
- EXTROSNB
- GENROSE
- LIARWHD
- MANCINO
- MODBEALE
- NONDIA
- NONMSQRT
- SBRYBND
- SROSENBR
- SSBRYBND
- TQUARTIC
-
reporter Thanks for sifting through this list already. It would be nice to have a way of associating a SUR2 problem and its feasibility counterpart. Some have a name ending in LS, some in NE, but many others follow a different pattern.
-
repo owner Agreed. Unfortunately many of these are historical, so we really shouldn't rename them as others may have already used them in comparisons.
-
reporter Right. They're the subset of the NOR2 problems that only have equality constraints. Unfortunately,
select
doesn't allow one to pick out problems with equality constraints only. But with thestats
driver, it's possible to find them. -
repo owner Missing variants now added. Issue closed.
Nick
-
repo owner - changed status to resolved
All problems mentioned now have equation equivalents
-
reporter Many thanks, Nick!!!
- Log in to comment
Agreed that some of these need equation variants.
Some comments, though
CURLY and SCURLY are misclassified I believe, they should be OUR2
Same for GENHUMPS and VAREIGVL
SENSORS is a negative sum of squares (and thus also misclassified)
OSCI* have equation variants OSCIPANE and OSCIGRNE FREUROTH has FREURONE
I will create feasibility variants when I get time, and correct the above errors.
Nick