PSD kernel densities give a convex MCAR imputer with closed-form marginals, KL consistency rates that adapt to smoothness, and competitive energy-distance performance on small real tables.
Explicit and recursive estimates of the Lambert W function
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
Solutions to a wide variety of transcendental equations can be expressed in terms of the Lambert $\mathrm{W}$ function. The $\mathrm{W}$ function, occurring frequently in applications, is a non-elementary, but now standard mathematical function implemented in all major technical computing systems. In this work, we discuss some approximations of the two real branches, $\mathrm{W}_0$ and $\mathrm{W}_{-1}$. On the one hand, we present some analytic lower and upper bounds on $\mathrm{W}_0$ for large arguments that improve on some earlier results in the literature. On the other hand, we analyze two logarithmic recursions, one with linear, and the other with quadratic rate of convergence. We propose suitable starting values for the recursion with quadratic rate that ensure convergence on the whole domain of definition of both real branches. We also provide a priori, simple, explicit and uniform estimates on its convergence speed that enable guaranteed, high-precision approximations of $\mathrm{W}_0$ and $\mathrm{W}_{-1}$ at any point. Finally, as an application of the $\mathrm{W}_0$ function, we settle a conjecture about the growth rate of the positive non-trivial solutions to the equation $x^y=y^x$.
fields
stat.ML 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation
PSD kernel densities give a convex MCAR imputer with closed-form marginals, KL consistency rates that adapt to smoothness, and competitive energy-distance performance on small real tables.