------------------------------------------------------------------------
Version 1.8
------------------------------------------------------------------------
2013-05-25 Ewout van den Berg <vandenberge.ewout@gmail.com>
* spg_mmv.m: Fixed linesearch bug.
------------------------------------------------------------------------
Version 1.7
------------------------------------------------------------------------
2009-05-20 Ewout van den Berg <ewout78@cs.ubc.ca>
* spg_mmv.m: Fixed bug in handling implicit matrices. Thanks to
Muhammad Usman for reporting this bug.
2008-09-18 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgSetParms.m: Replaced isstr by ischar.
------------------------------------------------------------------------
Version 1.6
------------------------------------------------------------------------
2008-05-02 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: The options structure now contains three fields to
specify routines for computing the primal norm, dual norm, and
for the Euclidean projection onto the associated ball.
* spg_mmv.m: New interface for multiple measurement vector case
of BP and BPDN.
* spg_group.m: New interface for group sparse BP and BPDN.
------------------------------------------------------------------------
Version 1.5
------------------------------------------------------------------------
2008-06-17 Michael P. Friedlander <mpf@cs.ubc.ca>
* spg_lasso.m, spg_bp.m, spg_bpdn.m: New interfaces. These are
specialized to specialized problem classes, and are simply
easy-to-use wrappers to spgl1.m.
* spgexamples.m: Added html documentation.
2008-05-02 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: Added support for class implementation of A.
------------------------------------------------------------------------
Version 1.4
------------------------------------------------------------------------
2008-03-17 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: Added isreal(x) check in projection code to ensure the
proper procedure is used even if realx is set incorrectly.
* spgl1.m: Check added to ensure tau is not updated in two
successive iterations. Thanks to Gilles Hennenfent for reporting
this problem.
2008-03-02 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgdemo.m: Added demo for basis pursuit with weighted one-norm
2008-02-23 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: Fixed bug in handling weighted one-norm ball orthogonal
projection. Thanks to Xiangrui Meng for reporting this bug.
* private/oneProjector.m: New one-projector wrapper.
* private/oneProjectorMex.c: New mex-interface to one-projector.
* private/oneProjectorCore.c: New implementation of (weighted)
one-norm ball orthogonal projection code.
------------------------------------------------------------------------
Version 1.3
------------------------------------------------------------------------
2007-12-04 Michael P. Friedlander <mpf@cs.ubc.ca>
* lsqr.m: Moved to "private" dir so that it won't interfere with
default lsqr function.
2007-12-03 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: Changed single tau stopping criterion 'f < optTol' to
'rNorm < optTol*bNorm'.
* spgl1.m: Implemented the code of dealing with the maximum number
of matrix-vector operations option.
2007-11-21 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: added the safeguard 'max(0,...)' to update for tau;
tau = max(0,tau + (rNorm * aError1) / gNorm);
This is needed to avoid tau from becoming negative, e.g., in
case sigma and the intial tau are large.
2007-11-30 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: EXIT_MATVEC_LIMIT exit condition added
* spgl1.m, spgSetParms.m: added .maxMatvec option
2007-11-26 Ewout van den Berg <ewout78@cs.ubc.ca>
* spgl1.m: added .rGap field in info
* ensure.m: added, used instead of the assert function
* spgl1.m, spgSetParms.m: added .iscomplex flag to options
2007-11-17 Michael P. Friedlander <mpf@cs.ubc.ca>
* spgl1.m: Default number of iterations did not match comment
string. Fixed. (Reported by Ozgur Yilmaz.)