Alexander Raichev and Mark C. Wilson, to appear in Proceedings of the International Conference on Analysis of Algorithms (Juan-les-Pins, June 2007).
Let $latex sum_{mathbf{n}inmathbb{N}^d} f_{mathbf{n}} mathbf{x}^mathbf{n}$ be a multivariate generating function that converges in a neighborhood of the origin of $latex mathbb{C}^d$. We present a new, multivariate method for computing the asymptotics of the diagonal coefficients $latex f_{a_1n,ldots,a_dn}$ and show its superiority over the standard, univariate diagonal method.
Note: there is a typo: in Example 3.6, we should have $latex c = ( (L-b)/a, (L-a)/b )$ where $latex L = sqrt{a^2 + b^2}$.