[SciPy-Dev] Introduction and feature proposal.

[Pauli Virtanen]

Christopher Miller <cmiller730 <at> gmail.com> writes:
[clip]
> p.s. Here are some references. Most of my work with these was in solving 
> stochastic PDEs, so the references tilt that way, but they are generally 
> useful tools for multi-dimensional integration/interpolation.
> 
> D. Xiu and J. Hesthaven. High-order collocation methods for differential 
> equa-
> tions with random inputs. SIAM Journal on Scientific Computing, 27:1118–
> 1139, 2005.
> 
> F. Nobile, R. Tempone, and C.G. Webster. An anisotropic sparse grid col-
> location algorithm for the solution of stochastic differential 
> equations. SIAM
> Journal on Numerical Analysis, 46:2411–2442, 2008.

Sparse grid quadratures would be a very welcome addition to 
scipy.integrate. Right now, the only N-dim integration routine
there is via nested 1-D integration, which is not efficient.

Adaptive sparse grids for sampling and interpolation in N dimensions
(of functions slow to evaluate and containing singular points) would
also be interesting for me; currently I'm sticking to ad-hoc
triangulation based methods.

Adapting existing Matlab/C code is usually reasonably
straightforward, feel free to ask if you want particular advice!

-- 
Pauli Virtanen