Searcharxiv⌕ Search

arXiv subjects

Xiao Fang

Publications and source records attributed to Xiao Fang.

At least 109 records · Page 6Linked to original sources
arXiv2017

A Practically Competitive and Provably Consistent Algorithm for Uplift Modeling

Randomized experiments have been critical tools of decision making for decades. However, subjects can show significant heterogeneity in response to treatments in many important applications. Therefore it is not enough to simply know which treatment is optimal for the entire population. What we need is a model that correctly customize treatment assignment base on subject characteristics. The problem of constructing such models from randomized experiments data is known as Uplift Modeling in the literature. Many algorithms have been proposed for uplift modeling and some have generated promising results on various data sets. Yet little is known about the theoretical properties of these algorithms. In this paper, we propose a new tree-based ensemble algorithm for uplift modeling. Experiments show that our algorithm can achieve competitive results on both synthetic and industry-provided data. In addition, by properly tuning the "node size" parameter, our algorithm is proved to be consistent under mild regularity conditions. This is the first consistent algorithm for uplift modeling that we are aware of.

cs.LG↗
arXiv2017

Beyond linear galaxy alignments

Galaxy intrinsic alignments (IA) are a critical uncertainty for current and future weak lensing measurements. We describe a perturbative expansion of IA, analogous to the treatment of galaxy biasing. From an astrophysical perspective, this model includes the expected large-scale alignment mechanisms for galaxies that are pressure-supported (tidal alignment) and rotation-supported (tidal torquing) as well as the cross-correlation between the two. Alternatively, this expansion can be viewed as an effective model capturing all relevant effects up to the given order. We include terms up to second order in the density and tidal fields and calculate the resulting IA contributions to two-point statistics at one-loop order. For fiducial amplitudes of the IA parameters, we find the quadratic alignment and linear-quadratic cross terms can contribute order-unity corrections to the total intrinsic alignment signal at $k\sim0.1\,h^{-1}{\rm Mpc}$, depending on the source redshift distribution. These contributions can lead to significant biases on inferred cosmological parameters in Stage IV photometric weak lensing surveys. We perform forecasts for an LSST-like survey, finding that use of the standard "NLA" model for intrinsic alignments cannot remove these large parameter biases, even when allowing for a more general redshift dependence. The model presented here will allow for more accurate and flexible IA treatment in weak lensing and combined probes analyses, and an implementation is made available as part of the public FAST-PT code. The model also provides a more advanced framework for understanding the underlying IA processes and their relationship to fundamental physics.

astro-ph.CO↗
arXiv2017

Quenching Measurements and Modeling of a Boron-Loaded Organic Liquid Scintillator

Organic liquid scintillators are used in a wide variety of applications in experimental nuclear and particle physics. Boron-loaded scintillators are particularly useful for detecting neutron captures, due to the high thermal neutron capture cross section of $^{10}$B. These scintillators are commonly used in neutron detectors, including the DarkSide-50 neutron veto, where the neutron may produce a signal when it scatters off protons in the scintillator or when it captures on $^{10}$B. Reconstructing the energy of these recoils is complicated by scintillation quenching. Understanding how nuclear recoils are quenched in these scintillators is an important and difficult problem. In this article, we present a set of measurements of neutron-induced proton recoils in a boron-loaded organic liquid scintillator at recoil energies ranging from 57--467 keV, and we compare these measurements to predictions from different quenching models. We find that a modified Birks' model whose denominator is quadratic in $dE/dx$ best describes the measurements, with $χ^2$/NDF$=1.6$. This result will help model nuclear recoil scintillation in similar detectors and can be used to improve their neutron tagging efficiency.

physics.ins-det↗
arXiv2017

Uplift Modeling with Multiple Treatments and General Response Types

Randomized experiments have been used to assist decision-making in many areas. They help people select the optimal treatment for the test population with certain statistical guarantee. However, subjects can show significant heterogeneity in response to treatments. The problem of customizing treatment assignment based on subject characteristics is known as uplift modeling, differential response analysis, or personalized treatment learning in literature. A key feature for uplift modeling is that the data is unlabeled. It is impossible to know whether the chosen treatment is optimal for an individual subject because response under alternative treatments is unobserved. This presents a challenge to both the training and the evaluation of uplift models. In this paper we describe how to obtain an unbiased estimate of the key performance metric of an uplift model, the expected response. We present a new uplift algorithm which creates a forest of randomized trees. The trees are built with a splitting criterion designed to directly optimize their uplift performance based on the proposed evaluation method. Both the evaluation method and the algorithm apply to arbitrary number of treatments and general response types. Experimental results on synthetic data and industry-provided data show that our algorithm leads to significant performance improvement over other applicable methods.

cs.AI↗
arXiv2017

FAST-PT: a novel algorithm to calculate convolution integrals in cosmological perturbation theory

We present a novel algorithm, FAST-PT, for performing convolution or mode-coupling integrals that appear in nonlinear cosmological perturbation theory. The algorithm uses several properties of gravitational structure formation -- the locality of the dark matter equations and the scale invariance of the problem -- as well as Fast Fourier Transforms to describe the input power spectrum as a superposition of power laws. This yields extremely fast performance, enabling mode-coupling integral computations fast enough to embed in Monte Carlo Markov Chain parameter estimation. We describe the algorithm and demonstrate its application to calculating nonlinear corrections to the matter power spectrum, including one-loop standard perturbation theory and the renormalization group approach. We also describe our public code (in Python) to implement this algorithm, including the applications described here.

astro-ph.CO↗
arXiv2017

The Application of SNiPER to the JUNO Simulation

JUNO is a multipurpose neutrino experiment which is designed to determine neutrino mass hierarchy and precisely measure oscillation parameters. As one of the important systems, the JUNO offline software is being developed using the SNiPER software. In this proceeding, we focus on the requirements of JUNO simulation and present the working solution based on the SNiPER. The JUNO simulation framework is in charge of managing event data, detector geometries and materials, physics processes, simulation truth information etc. It glues physics generator, detector simulation and electronics simulation modules together to achieve a full simulation chain. In the implementation of the framework, many attractive characteristics of the SNiPER have been used, such as dynamic loading, flexible flow control, multiple event management and Python binding. Furthermore, additional efforts have been made to make both detector and electronics simulation flexible enough to accommodate and optimize different detector designs. For the Geant4-based detector simulation, each sub-detector component is implemented as a SNiPER tool which is a dynamically loadable and configurable plugin. So it is possible to select the detector configuration at runtime. The framework provides the event loop to drive the detector simulation and interacts with the Geant4 which is implemented as a passive service. All levels of user actions are wrapped into different customizable tools, so that user functions can be easily extended by just adding new tools. The electronics simulation has been implemented by following an event driven scheme. The SNiPER task component is used to simulate data processing steps in the electronics modules. The electronics and trigger are synchronized by triggered events containing possible physics signals.

physics.ins-det↗
arXiv2017

FAST-PT II: an algorithm to calculate convolution integrals of general tensor quantities in cosmological perturbation theory

Cosmological perturbation theory is a powerful tool to predict the statistics of large-scale structure in the weakly non-linear regime, but even at 1-loop order it results in computationally expensive mode-coupling integrals. Here we present a fast algorithm for computing 1-loop power spectra of quantities that depend on the observer's orientation, thereby generalizing the FAST-PT framework (McEwen et al., 2016) that was originally developed for scalars such as the matter density. This algorithm works for an arbitrary input power spectrum and substantially reduces the time required for numerical evaluation. We apply the algorithm to four examples: intrinsic alignments of galaxies in the tidal torque model; the Ostriker-Vishniac effect; the secondary CMB polarization due to baryon flows; and the 1-loop matter power spectrum in redshift space. Code implementing this algorithm and these applications is publicly available at https://github.com/JoeMcEwen/FAST-PT .

astro-ph.CO↗
arXiv2016

A new probe of magnetic fields in the pre-reionization epoch: II. Detectability

In the first paper of this series, we proposed a novel method to probe large-scale intergalactic magnetic fields during the cosmic Dark Ages, using 21-cm tomography. This method relies on the effect of spin alignment of hydrogen atoms in a cosmological setting, and on the effect of magnetic precession of the atoms on the statistics of the 21-cm brightness-temperature fluctuations. In this paper, we forecast the sensitivity of future tomographic surveys to detecting magnetic fields using this method. For this purpose, we develop a minimum-variance estimator formalism to capture the characteristic anisotropy signal using the two-point statistics of the brightness-temperature fluctuations. We find that, depending on the reionization history, and subject to the control of systematics from foreground subtraction, an array of dipole antennas in a compact-grid configuration with a collecting area slightly exceeding one square kilometer can achieve a $1σ$ detection of $\sim$$10^{-21}$ Gauss comoving (scaled to present-day value) within three years of observation. Using this method, tomographic 21-cm surveys could thus probe ten orders of magnitude below current CMB constraints on primordial magnetic fields, and provide exquisite sensitivity to large-scale magnetic fields in situ at high redshift.

astro-ph.CO↗
arXiv2016

Utility-based Link Recommendation for Online Social Networks

Link recommendation, which suggests links to connect currently unlinked users, is a key functionality offered by major online social networks. Salient examples of link recommendation include "People You May Know" on Facebook and LinkedIn as well as "You May Know" on Google+. The main stakeholders of an online social network include users (e.g., Facebook users) who use the network to socialize with other users and an operator (e.g., Facebook Inc.) that establishes and operates the network for its own benefit (e.g., revenue). Existing link recommendation methods recommend links that are likely to be established by users but overlook the benefit a recommended link could bring to an operator. To address this gap, we define the utility of recommending a link and formulate a new research problem - the utility-based link recommendation problem. We then propose a novel utility-based link recommendation method that recommends links based on the value, cost, and linkage likelihood of a link, in contrast to existing link recommendation methods which focus solely on linkage likelihood. Specifically, our method models the dependency relationship between value, cost, linkage likelihood and utility-based link recommendation decision using a Bayesian network, predicts the probability of recommending a link with the Bayesian network, and recommends links with the highest probabilities. Using data obtained from a major U.S. online social network, we demonstrate significant performance improvement achieved by our method compared to prevalent link recommendation methods from representative prior research.

cs.SI↗
arXiv2015

A Survey of Link Recommendation for Social Networks: Methods, Theoretical Foundations, and Future Research Directions

Link recommendation has attracted significant attentions from both industry practitioners and academic researchers. In industry, link recommendation has become a standard and most important feature in online social networks, prominent examples of which include "People You May Know" on LinkedIn and "You May Know" on Google+. In academia, link recommendation has been and remains a highly active research area. This paper surveys state-of-the-art link recommendation methods, which can be broadly categorized into learning-based methods and proximity-based methods. We further identify social and economic theories, such as social interaction theory, that underlie these methods and explain from a theoretical perspective why a link recommendation method works. Finally, we propose to extend link recommendation research in several directions that include utility-based link recommendation, diversity of link recommendation, link recommendation from incomplete data, and experimental study of link recommendation.

cs.SI↗
arXiv2015

Poisson approximation for two scan statistics with rates of convergence

As an application of Stein's method for Poisson approximation, we prove rates of convergence for the tail probabilities of two scan statistics that have been suggested for detecting local signals in sequences of independent random variables subject to possible change-points. Our formulation deals simultaneously with ordinary and with large deviations.

math.PR↗
arXiv2015

Multivariate Normal Approximation by Stein's Method: The Concentration Inequality Approach

The concentration inequality approach for normal approximation by Stein's method is generalized to the multivariate setting. We use this approach to prove a non-smooth function distance for multivariate normal approximation for standardized sums of $k$-dimensional independent random vectors $W=\sum_{i=1}^n X_i$ with an error bound of order $k^{1/2}γ$ where $γ=\sum_{i=1}^n E|X_i|^3$. For sums of locally dependent (unbounded) random vectors, we obtain a fourth moment bound which is typically of order $O_k(1/\sqrt{n})$, as well as a third moment bound which is typically of order $O_k(\log n/\sqrt{n})$.

math.PR↗
arXiv2015

A multivariate CLT for bounded decomposable random vectors with the best known rate

We prove a multivariate central limit theorem with explicit error bound on a non-smooth function distance for sums of bounded decomposable $d$-dimensional random vectors. The decomposition structure is similar to that of Barbour, Karoński and Ruciński (1989) and is more general than the local dependence structure considered in Chen and Shao (2004). The error bound is of the order $d^{\frac{1}{4}} n^{-\frac{1}{2}}$, where $d$ is the dimension and $n$ is the number of summands. The dependence on $d$, namely $d^{\frac{1}{4}}$, is the best known dependence even for sums of independent and identically distributed random vectors, and the dependence on $n$, namely $n^{-\frac{1}{2}}$, is optimal. We apply our main result to a random graph example.

math.PR↗
arXiv2015

On the error bound in a combinatorial central limit theorem

Let $\mathbb{X}=\{X_{ij}: 1\le i,j\le n\}$ be an $n\times n$ array of independent random variables where $n\ge2$. Let $π$ be a uniform random permutation of $\{1,2,\dots,n\}$, independent of $\mathbb{X}$, and let $W=\sum_{i=1}^nX_{iπ(i)}$. Suppose $\mathbb{X}$ is standardized so that ${\mathbb{E}}W=0,\operatorname {Var}(W)=1$. We prove that the Kolmogorov distance between the distribution of $W$ and the standard normal distribution is bounded by $451\sum_{i,j=1}^n{\mathbb{E}}|X_{ij}|^3/n$. Our approach is by Stein's method of exchangeable pairs and the use of a concentration inequality.

math.PR↗
arXiv2015

Scintillation efficiency measurement of Na recoils in NaI(Tl) below the DAMA/LIBRA energy threshold

The dark matter interpretation of the DAMA modulation signal depends on the NaI(Tl) scintillation efficiency of nuclear recoils. Previous measurements for Na recoils have large discrepancies, especially in the DAMA/LIBRA modulation energy region. We report a quenching effect measurement of Na recoils in NaI(Tl) from 3keV$_{\text{nr}}$ to 52keV$_{\text{nr}}$, covering the whole DAMA/LIBRA energy region for light WIMP interpretations. By using a low-energy, pulsed neutron beam, a double time-of-flight technique, and pulse-shape discrimination methods, we obtained the most accurate measurement of this kind for NaI(Tl) to date. The results differ significantly from the DAMA reported values at low energies, but fall between the other previous measurements. We present the implications of the new quenching results for the dark matter interpretation of the DAMA modulation signal.

physics.ins-det↗
arXiv2014

A universal error bound in the CLT for counting monochromatic edges in uniformly colored graphs

Let $\{G_n: n\geq 1\}$ be a sequence of simple graphs. Suppose $G_n$ has $m_n$ edges and each vertex of $G_n$ is colored independently and uniformly at random with $c_n$ colors. Recently, Bhattacharya, Diaconis and Mukherjee (2013) proved universal limit theorems for the number of monochromatic edges in $G_n$. Their proof was by the method of moments, and therefore was not able to produce rates of convergence. By a non-trivial application of Stein's method, we prove that there exists a universal error bound for their central limit theorem. The error bound depends only on $m_n$ and $c_n$, regardless of the graph structure.

math.PR↗
arXiv2014

Discretized normal approximation by Stein's method

We prove a general theorem to bound the total variation distance between the distribution of an integer valued random variable of interest and an appropriate discretized normal distribution. We apply the theorem to 2-runs in a sequence of i.i.d. Bernoulli random variables, the number of vertices with a given degree in the Erdös-Rényi random graph, and the uniform multinomial occupancy model.

math.PR↗