mp.polyroots could use fp.polyroots to generate initial values. This would
improve speed quite significantly.
Further, it should handle multiple roots better. The Durand-Kerner method
can be modified to detect and count multiple roots; see
http://portal.acm.org/citation.cfm?id=117554
The multiple roots could be isolated and then refined to full precision
using findroot with the appropriate algorithm.
Original issue reported on code.google.com by fredrik....@gmail.com on 12 Jan 2010 at 5:20
Original issue reported on code.google.com by
fredrik....@gmail.com
on 12 Jan 2010 at 5:20