… fmin_powell, fmin_ncg, fmin_l_bfgs_b. The fmin_l_bfgs_b function is currently broken and fails. (The problem may be in the handling of character arrays in f2py or lbfgsb.pyf; see http://www.scipy.net/pipermail/scipy-dev/2005-October/003685.html.)
….h" so Python.h is imported first.
…ith a length.
…item__ Reverted previous getnnz() behaviour.
* Improved robustness in an incredibly weird corner-case: when the interpolation point is on an edge (within FP error) and is *on* the edge of another circumcircle. I haven't figured out why it was wrong, but it was very easy to work around. * Added interpolation to unstructured sets of (x,y) points instead of just regular grids. * Added some unit tests. * Added extrapolation scheme. It shouldn't really be trusted, but it gets the job done. * Implemented some of the accuracy tests from Renka's TOMS 792 paper.
* fixed csrmucsc() bug in sparsetools wrapper * fixed bug in __rsub__ * enabled __radd__ unit tests * added functionality for dense * sparse multiplication * added more unit tests for multiplication * simplified csc_matrix.__init__() argument parsing * renamed conjtranspose() to getH() for compatibility with dense matrices * more code cleanups