Due to popular demand ;-) I added a maximum number of steps to the regression procedure. It seems that I didn't explain the "almost always" part of my previous email. The loop will *always* exit, but there's a chance that it will exit with an error (singular matrix). Regardless, it's good to have a maximum number of steps. Eventually we should let the user modify this value. I also realized that cut-off DELTA that I picked was unnecessarily small. An updated file is attached. Thanks for the input. Daniel.
Attachment:
nonlinear-regression.c
Description: update