有限离散情形下$ϕ$-投影的可微性
On the differentiability of $ϕ$-projections in the discrete finite case
AI总结:
针对有限空间有限测度情形,本文给出$ϕ$-投影连续可微的条件,在凸集等场景下简化验证,并将结果应用于$ϕ$-投影估计量,推导其稳健统计计算及多场景渐近性质。
摘要:
针对有限空间上的有限测度,我们给出了$ϕ$-投影具有连续可微性的条件。当$ϕ$-投影的目标集合为凸集时,我们证明所需假设可由易于验证的条件推导得出。特别地,对于输入概率向量及一大类$ϕ$-散度,当投影到由线性等式定义的集合时,$ϕ$-投影具有连续可微性。所得结果应用于$ϕ$-投影估计量(即最小$ϕ$-散度估计量):第一个应用源于稳健统计,涉及此类估计量影响函数的计算;第二个应用是推导其渐近性质,包括投影到概率向量参数集、固定矩分布生成的概率向量集及二元概率数组Fréchet类的情形,且无论被投影元素是否属于目标集合,渐近性质均成立。
英文摘要:
In the case of finite measures on finite spaces, we state conditions under which ϕ- projections are continuously differentiable. When the set on which one wishes to ϕ- project is convex, we show that the required assumptions are implied by easily verifiable conditions. In particular, for input probability vectors and a rather large class of ϕ-divergences, we obtain that ϕ-projections are continuously differentiable when projecting on a set defined by linear equalities. The obtained results are applied to ϕ- projection estimators (that is, minimum ϕ-divergence estimators). A first application, rooted in robust statistics, concerns the computation of the influence functions of such estimators. In a second set of applications, we derive their asymptotics when projecting on parametric sets of probability vectors, on sets of probability vectors generated from distributions with certain moments fixed and on Fréchet classes of bivariate probability arrays. The resulting asymptotics hold whether the element to be ϕ-projected belongs to the set on which one wishes to ϕ-project or not.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。