Lipschitz Density-Ratios, Structured Data, and Data-driven Tuning
نویسنده
چکیده
Density-ratio estimation (i.e. estimating f = fQ/fP for two unknown distributions Q and P ) has proved useful in many Machine Learning tasks, e.g., risk-calibration in transfer-learning, two-sample tests, and also useful in common techniques such importance sampling and bias correction. While there are many important analyses of this estimation problem, the present paper derives convergence rates in other practical settings that are less understood, namely, extensions of traditional Lipschitz smoothness conditions, and common high-dimensional settings with structured data (e.g. manifold data, sparse data). Various interesting facts, which hold in earlier settings, are shown to extend to these settings. Namely, (1) optimal rates depend only on the smoothness of the ratio f , and not on the densities fQ, fP , supporting the belief that plugging in estimates for fQ, fP is suboptimal; (2) optimal rates depend only on the intrinsic dimension of data, i.e. this problem – unlike density estimation – escapes the curse of dimension. We further show that near-optimal rates are attainable by estimators tuned from data alone, i.e. with no prior distributional information. This last fact is of special interest in unsupervised settings such as this one, where only oracle rates seem to be known, i.e., rates which assume critical distributional information usually unavailable in practice.
منابع مشابه
تأثیر ساختواژهها در تجزیه وابستگی زبان فارسی
Data-driven systems can be adapted to different languages and domains easily. Using this trend in dependency parsing was lead to introduce data-driven approaches. Existence of appreciate corpora that contain sentences and theirs associated dependency trees are the only pre-requirement in data-driven approaches. Despite obtaining high accurate results for dependency parsing task in English langu...
متن کاملA novel existence and uniqueness theorem for solutions to FDEs driven by Lius process with weak Lipschitz coefficients
This paper we investigate the existence and uniqueness of solutions to fuzzydierential equations driven by Liu's process. For this, it is necessary to provideand prove a new existence and uniqueness theorem for fuzzy dierential equationsunder weak Lipschitz condition. Then the results allows us to considerand analyze solutions to a wide range of nonlinear fuzzy dierential equationsdriven by Liu...
متن کاملLipschitz Equivalence of Cantor Sets and Algebraic Properties of Contraction Ratios
In this paper we investigate the Lipschitz equivalence of dust-like self-similar sets in Rd. One of the fundamental results by Falconer and Marsh [On the Lipschitz equivalence of Cantor sets, Mathematika, 39 (1992), 223– 233] establishes conditions for Lipschitz equivalence based on the algebraic properties of the contraction ratios of the self-similar sets. In this paper we extend the study by...
متن کاملForecasting Ozone Density in Tehran Air Using a Smart Data-Driven Approach
Introduction: As a metropolitan area in Iran, Tehran is exposed to damage from air pollution due to its large population and pollutants from various sources. Accordingly, research on damage induced by air pollution in this city seems necessary. The main purpose of this study was to forecast ozone in the city of Tehran. Considering the hazards of ozone (O3) gas on human health and the environmen...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2017