Added comments.

pampelmuse/core/fit_layer.py

@@ -18,12 +18,46 @@ GNU General Public License for more details.
 You should have received a copy of the GNU General Public License
 along with PampelMuse. If not, see <http://www.gnu.org/licenses/>.
 
+
 +++
 This module implements the optimization for a single layer of IFS data.
 
+The optimization consists of three steps that are performed iteratively:
+(1) An optimization for the fluxes based on the current estimate of the PSF
+    parameters and the source coordinates. This is a linear optimization,
+    carried out via direct inversion of a sparse matrix that contains the PSF
+    profiles of all fitted components
+(2) An optimization for the positions of the individual sources that is
+    performed using a Levenberg-Marquard algorithm. Note that there are two
+    different ways of how the to fit the fluxes.
+    (a) via a coordinate transformation that translates the positions from a
+        reference frame into the image that is being analysed. In that case,
+        the parameters of the transformation are optimized, not individual
+        source positions.
+    (b) via direct coordinate fits without a transformation. This approach is
+        currently only recommended for sparsely populated fields.
+(3) An optimization for the parameters of the PSF profile that is also
+    performed using a Levenberg-Marquard algorithm.
+
+Note that steps (2a) and (3) can be performed in two different ways, either by
+simultaneously fitting all sources or by selecting individual isolated sources
+and only fitting this subset. The former approach is recommended for small
+images while the latter approach works better on large images. Step (2b) only
+works with the latter approach.
+
+Finally, a not about the fitting weights of the individual pixels in the
+analysis. By default, inverse-variance weighting is used. If the variances are
+not provided, the code tries to estimate them from the current residuals as a
+function of the fitted pixel intensities.
+In addition, the code uses an additional weighting scheme to reduce the
+weights of deviant pixels. As this method requires a reasonably good model to
+stars with, it is currently only used if the source positions and PSF
+parameters are not fitted.
+
+
 Latest SVN revision
 -------------------
-349, 2016/08/05
+351, 2016/08/07
 """
 import contextlib
 import logging
@@ -36,7 +70,7 @@ from ..core.coordinates import Transformation
 
 
 __author__ = "Sebastian Kamann (skamann@astro.physik.uni-goettingen.de)"
-__revision__ = 349
+__revision__ = 351
 
 
 logger = logging.getLogger(__name__)