arXiv subjects

Xi Chen

Publications and source records attributed to Xi Chen.

At least 1,045 records · Page 58Linked to original sources

arXiv2018

Thermodynamics of Spin-1/2 Kagomé Heisenberg Antiferromagnet: Algebraic Paramagnetic Liquid and Finite-Temperature Phase Diagram

Quantum fluctuations from frustration can trigger quantum spin liquids (QSLs) at zero temperature. However, it is unclear how thermal fluctuations affect a QSL. We employ state-of-the-art tensor network-based methods to explore the ground state and thermodynamic properties of the spin-1/2 kagome Heisenberg antiferromagnet (KHA). Its ground state is shown to be consistent with a gapless QSL by observing the absence of zero-magnetization plateau as well as the algebraic behaviors of susceptibility and specific heat at low temperatures, respectively. We show that there exists an \textit{algebraic paramagnetic liquid} (APL) that possesses both the paramagnetic properties and the algebraic behaviors inherited from the QSL. The APL is induced under the interplay between quantum fluctuations from geometrical frustration and thermal fluctuations. By studying the temperature-dependent behaviors of specific heat and magnetic susceptibility, a finite-temperature phase diagram in a magnetic field is suggested, where various phases are identified. This present study gains useful insight into the thermodynamic properties of the spin-1/2 KHA with or without a magnetic field and is helpful for relevant experimental studies.

cond-mat.str-el↗

arXiv2018

Deep Learning by Doing: The NVIDIA Deep Learning Institute and University Ambassador Program

Over the past two decades, High-Performance Computing (HPC) communities have developed many models for delivering education aiming to help students understand and harness the power of parallel and distributed computing. Most of these courses either lack a hands-on component or heavily focus on theoretical characterization behind complex algorithms. To bridge the gap between application and scientific theory, NVIDIA Deep Learning Institute (DLI) (www.nvidia.com/dli) has designed an on-line education and training platform that helps students, developers, and engineers solve real-world problems in a wide range of domains using deep learning and accelerated computing. DLI's accelerated computing course content starts with the fundamentals of accelerating applications with CUDA and OpenACC in addition to other courses in training and deploying neural networks for deep learning. Advanced and domain-specific courses in deep learning are also available. The online platform enables students to use the latest AI frameworks, SDKs, and GPU-accelerated technologies on fully-configured GPU servers in the cloud so the focus is more on learning and less on environment setup. Students are offered project-based assessment and certification at the end of some courses. To support academics and university researchers teaching accelerated computing and deep learning, the DLI University Ambassador Program enables educators to teach free DLI courses to university students, faculty, and researchers.

cs.CY↗

arXiv2018

Parallelized Interactive Machine Learning on Autonomous Vehicles

Deep reinforcement learning (deep RL) has achieved superior performance in complex sequential tasks by learning directly from image input. A deep neural network is used as a function approximator and requires no specific state information. However, one drawback of using only images as input is that this approach requires a prohibitively large amount of training time and data for the model to learn the state feature representation and approach reasonable performance. This is not feasible in real-world applications, especially when the data are expansive and training phase could introduce disasters that affect human safety. In this work, we use a human demonstration approach to speed up training for learning features and use the resulting pre-trained model to replace the neural network in the deep RL Deep Q-Network (DQN), followed by human interaction to further refine the model. We empirically evaluate our approach by using only a human demonstration model and modified DQN with human demonstration model included in the Microsoft AirSim car simulator. Our results show that (1) pre-training with human demonstration in a supervised learning approach is better and much faster at discovering features than DQN alone, (2) initializing the DQN with a pre-trained model provides a significant improvement in training time and performance even with limited human demonstration, and (3) providing the ability for humans to supply suggestions during DQN training can speed up the network's convergence on an optimal policy, as well as allow it to learn more complex policies that are harder to discover by random exploration.

cs.LG↗

arXiv2018

Pan-Cancer Epigenetic Biomarker Selection from Blood Samples Using SAS

A key focus in current cancer research is the discovery of cancer biomarkers that allow earlier detection with high accuracy and lower costs for both patients and hospitals. Blood samples have long been used as a health status indicator, but DNA methylation signatures in blood have not been fully appreciated in cancer research. Historically, analysis of cancer has been conducted directly with the patient's tumor or related tissues. Such analyses allow physicians to diagnose a patient's health and cancer status; however, physicians must observe certain symptoms that prompt them to use biopsies or imaging to verify the diagnosis. This is a post-hoc approach. Our study will focus on epigenetic information for cancer detection, specifically information about DNA methylation in human peripheral blood samples in cancer discordant monozygotic twin-pairs. This information might be able to help us detect cancer much earlier, before the first symptom appears. Several other types of epigenetic data can also be used, but here we demonstrate the potential of blood DNA methylation data as a biomarker for pan-cancer using SAS 9.3 and SAS EM. We report that 55 methylation CpG sites measurable in blood samples can be used as biomarkers for early cancer detection and classification.

q-bio.GN↗

arXiv2018

A Method to Facilitate Cancer Detection and Type Classification from Gene Expression Data using a Deep Autoencoder and Neural Network

With the increased affordability and availability of whole-genome sequencing, large-scale and high-throughput gene expression is widely used to characterize diseases, including cancers. However, establishing specificity in cancer diagnosis using gene expression data continues to pose challenges due to the high dimensionality and complexity of the data. Here we present models of deep learning (DL) and apply them to gene expression data for the diagnosis and categorization of cancer. In this study, we have developed two DL models using messenger ribonucleic acid (mRNA) datasets available from the Genomic Data Commons repository. Our models achieved 98% accuracy in cancer detection, with false negative and false positive rates below 1.7%. In our results, we demonstrated that 18 out of 32 cancer-typing classifications achieved more than 90% accuracy. Due to the limitation of a small sample size (less than 50 observations), certain cancers could not achieve a higher accuracy in typing classification, but still achieved high accuracy for the cancer detection task. To validate our models, we compared them with traditional statistical models. The main advantage of our models over traditional cancer detection is the ability to use data from various cancer types to automatically form features to enhance the detection and diagnosis of a specific cancer type.

stat.ML↗

arXiv2018

Modeling Personalized Dynamics of Social Network and Opinion at Individual Level

Network dynamics has always been a meaningful topic deserving exploration in the realm of academy. previous network models contain two parts: (1) generating structure as per user property; (2) changing property as per network structure. Properties in these models, however, cannot be interpreted to concept in prevalent social theories or empirical truth. Also, they usually treat everyone in an uniform fashion. While such assumption is quite misguiding, and thus saliently limits their performance. To overcome these flaws, we devise a personalized evolving model for social network and opinion (PENO), where citizens' ideology is revealed by variable opinions and four dimensions of personality are considered for each entity - leadership, openness, agreeableness, and neuroticism. Opinion propagates via social tie, tie generates from opinion affinity, and personalities integrally work with opinion and tie across evolution. To our best knowledge, PENO is the first attempt to introduce personality impact in network dynamics and verify social science with reasonable visualization during simulation. We also present its probabilistic graph and conceive iterative learning algorithm. Experiments show PENO outperforms several state-of-the-art baselines over two typical prediction tasks - congress voting prediction for legislative bills and friendship prediction on a book-commenting website. Finally, we discuss its scalability to do multi-task learning and transfer learning in daily scenarios.

cs.SI↗

arXiv2018

Updated Core Libraries of the ALPS Project

The open source ALPS (Algorithms and Libraries for Physics Simulations) project provides a collection of physics libraries and applications, with a focus on simulations of lattice models and strongly correlated electron systems. The libraries provide a convenient set of well-documented and reusable components for developing condensed matter physics simulation codes, and the applications strive to make commonly used and proven computational algorithms available to a non-expert community. In this paper we present an update of the core ALPS libraries. We present in particular new Monte Carlo libraries and new Green's function libraries.

physics.comp-ph↗

arXiv2018

Improving the efficiency and robustness of nested sampling using posterior repartitioning

In real-world Bayesian inference applications, prior assumptions regarding the parameters of interest may be unrepresentative of their actual values for a given dataset. In particular, if the likelihood is concentrated far out in the wings of the assumed prior distribution, this can lead to extremely inefficient exploration of the resulting posterior by nested sampling algorithms, with unnecessarily high associated computational costs. Simple solutions such as broadening the prior range in such cases might not be appropriate or possible in real-world applications, for example when one wishes to assume a single standardised prior across the analysis of a large number of datasets for which the true values of the parameters of interest may vary. This work therefore introduces a posterior repartitioning (PR) method for nested sampling algorithms, which addresses the problem by redefining the likelihood and prior while keeping their product fixed, so that the posterior inferences and evidence estimates remain unchanged but the efficiency of the nested sampling process is significantly increased. Numerical results show that the PR method provides a simple yet powerful refinement for nested sampling algorithms to address the issue of unrepresentative priors.

stat.CO↗

arXiv2018

Uniqueness of Transformation based on Jacobian Determinant and curl-Vector

Numerical examples demonstrated that a prescribed positive Jacobian determinant alone can not uniquely determine a diffeomorphism. It is conjectured that the uniqueness of a transformation can be assured by its Jacobian determinant and the curl-vector. In this work, we study the uniqueness problem analytically and propose an approach to the proof of the uniqueness of a transformation with prescribed Jacobian determinant and curl-vector.

cs.CG↗

arXiv2018

Computational Technologies for Brain Morphometry

In this paper, we described a set of computational technologies for image analysis with applications in Brain Morphometry. The proposed technologies are based on a new Variational Principle which constructs a transformation with prescribed Jacobian determinant (which models local size changes) and prescribed curl-vector (which models local rotations). The goal of this research is to convince the image research community that Jacobian determinant as well as curl-vector should be used in all steps of image analysis. Specifically, we develop an optimal control method for non-rigid registration; a new concept and construction of average transformation; and a general robust method for construction of unbiased template from a set of images. Computational examples are presented to show the effects of curl-vector and the effectiveness of optimal control methods for non-rigid registration and our method for construction of unbiased template.

cs.CG↗

arXiv2018

Quantile Regression Under Memory Constraint

This paper studies the inference problem in quantile regression (QR) for a large sample size $n$ but under a limited memory constraint, where the memory can only store a small batch of data of size $m$. A natural method is the naïve divide-and-conquer approach, which splits data into batches of size $m$, computes the local QR estimator for each batch, and then aggregates the estimators via averaging. However, this method only works when $n=o(m^2)$ and is computationally expensive. This paper proposes a computationally efficient method, which only requires an initial QR estimator on a small batch of data and then successively refines the estimator via multiple rounds of aggregations. Theoretically, as long as $n$ grows polynomially in $m$, we establish the asymptotic normality for the obtained estimator and show that our estimator with only a few rounds of aggregations achieves the same efficiency as the QR estimator computed on all the data. Moreover, our result allows the case that the dimensionality $p$ goes to infinity. The proposed method can also be applied to address the QR problem under distributed computing environment (e.g., in a large-scale sensor network) or for real-time streaming data.

stat.ME↗

arXiv2018

Detection of 84-GHz class I methanol maser emission towards NGC 253

We have investigated the central region of NGC 253 for the presence of 84.5-GHz ($5_{-1}\rightarrow4_0$E) methanol emission using the Australia Telescope Compact Array. We present the second detection of 84.5-GHz class~I methanol maser emission outside the Milky Way. This maser emission is offset from dynamical centre of NGC 253, in a region with previously detected emission from class~I maser transitions (36.2-GHz $4_{-1}\rightarrow3_0$E and 44.1-GHz $7_{0}\rightarrow6_1$A$^{+}$ methanol lines) . The emission features a narrow linewidth ($\sim$12 km s$^{-1}$) with a luminosity approximately 5 orders of magnitude higher than typical Galactic sources. We determine an integrated line intensity ratio of $1.2\pm0.4$ between the 36.2 GHz and 84.5-GHz class I methanol maser emission, which is similar to the ratio observed towards Galactic sources. The three methanol maser transitions observed toward NGC 253 each show a different distribution, suggesting differing physical conditions between the maser sites and that observations of additional class~I methanol transitions will facilitate investigations of the maser pumping regime.

astro-ph.GA↗

arXiv2018

Equivalence Between Policy Gradients and Soft Q-Learning

Two of the leading approaches for model-free reinforcement learning are policy gradient methods and $Q$-learning methods. $Q$-learning methods can be effective and sample-efficient when they work, however, it is not well-understood why they work, since empirically, the $Q$-values they estimate are very inaccurate. A partial explanation may be that $Q$-learning methods are secretly implementing policy gradient updates: we show that there is a precise equivalence between $Q$-learning and policy gradient methods in the setting of entropy-regularized reinforcement learning, that "soft" (entropy-regularized) $Q$-learning is exactly equivalent to a policy gradient method. We also point out a connection between $Q$-learning methods and natural policy gradient methods. Experimentally, we explore the entropy-regularized versions of $Q$-learning and policy gradients, and we find them to perform as well as (or slightly better than) the standard variants on the Atari benchmark. We also show that the equivalence holds in practical settings by constructing a $Q$-learning method that closely matches the learning dynamics of A3C without using a target network or $ε$-greedy exploration schedule.

cs.LG↗

arXiv2018

Fast long-range state transfer in quantum dot arrays via shortcuts to adiabaticity

Rapid and efficient preparation, manipulation and transfer of quantum states through an array of quantum dots (QDs) is a demanding requisite task for quantum information processing and quantum computation in solid-state physics. Conventional adiabatic protocols, as coherent transfer by adiabatic passage (CTAP) and its variations, provide slow transfer prone to decoherence, which could lower the fidelity to some extent. To achieve the robustness against decoherence, we propose a protocol of speeding up the adiabatic charge transfer in multi-QD systems, sharing the concept of "Shortcuts to Adiabaticity" (STA). We first apply the STA techniques, including the counterdiabatic driving and inverse engineering, to speed up the direct (long range) transfer between edge dots in triple QDs. Then, we extend our analysis to a multi-dot system. We show how by implementing the modified pulses, fast adiabatic-like charge transport between the outer dots can be eventually achieved without populating intermediate dots. We discuss as well the dependence of the transfer fidelity on the operation time in the presence of dephasing. The proposed protocols for accelerating adiabatic charge transfer directly between the outer dots in a QD array offers a robust mechanism for quantum information processing, by minimizing decoherence and relaxation processes.

cond-mat.mes-hall↗

arXiv2018

Query Tracking for E-commerce Conversational Search: A Machine Comprehension Perspective

With the development of dialog techniques, conversational search has attracted more and more attention as it enables users to interact with the search engine in a natural and efficient manner. However, comparing with the natural language understanding in traditional task-oriented dialog which focuses on slot filling and tracking, the query understanding in E-commerce conversational search is quite different and more challenging due to more diverse user expressions and complex intentions. In this work, we define the real-world problem of query tracking in E-commerce conversational search, in which the goal is to update the internal query after each round of interaction. We also propose a self attention based neural network to handle the task in a machine comprehension perspective. Further more we build a novel E-commerce query tracking dataset from an operational E-commerce Search Engine, and experimental results on this dataset suggest that our proposed model outperforms several baseline methods by a substantial gain for Exact Match accuracy and F1 score, showing the potential of machine comprehension like model for this task.

cs.CL↗

arXiv2018

On Degrees of Freedom of Projection Estimators with Applications to Multivariate Nonparametric Regression

In this paper, we consider the nonparametric regression problem with multivariate predictors. We provide a characterization of the degrees of freedom and divergence for estimators of the unknown regression function, which are obtained as outputs of linearly constrained quadratic optimization procedures, namely, minimizers of the least squares criterion with linear constraints and/or quadratic penalties. As special cases of our results, we derive explicit expressions for the degrees of freedom in many nonparametric regression problems, e.g., bounded isotonic regression, multivariate (penalized) convex regression, and additive total variation regularization. Our theory also yields, as special cases, known results on the degrees of freedom of many well-studied estimators in the statistics literature, such as ridge regression, Lasso and generalized Lasso. Our results can be readily used to choose the tuning parameter(s) involved in the estimation procedure by minimizing the Stein's unbiased risk estimate. As a by-product of our analysis we derive an interesting connection between bounded isotonic regression and isotonic regression on a general partially ordered set, which is of independent interest.

math.ST↗

arXiv2018

Target Search of a Protein on DNA in the Presence of Position-dependent Bias

We study the target searching on the DNA for proteins in the presence of non-constant drift and non-Gaussian $α$-stable Lévy fluctuations. The target searching is realized by the facilitated diffusion process. The existing works are about this problem in the case of constant drift. Starting from a non-local Fokker-Planck equation with a "sink" term, we obtain the possibility density function for the protein occurring at position $x$ on time $t$. Based on this, we further compute the survival probability and the first arrival density in order to quantify the searching mechanisms. The numerical results show that in the linear drift case, there is an optimal $α$ index for the search to be most likely successful (searching reliability reaches its maximum). This optimal $α$ index depends on initial position-target separation. It is also found that the diffusion intensity plays a positive role in improving the searching success. The nonlinear double-well drift could drive the protein to reach the target with a larger possibility than the linear drag at initial time period, but viewing at the long time evolution, the linear drift is more beneficial for target searching success. In contrast to the linear drift case, the search reliability and efficiency with nonlinear drift have a monotonic relationship with the $α$ index, that is, the smaller the $α$ index is, the more possibly a protein finds its target.

cond-mat.stat-mech↗

arXiv2018

A Note on a Tight Lower Bound for MNL-Bandit Assortment Selection Models

In this short note we consider a dynamic assortment planning problem under the capacitated multinomial logit (MNL) bandit model. We prove a tight lower bound on the accumulated regret that matches existing regret upper bounds for all parameters (time horizon $T$, number of items $N$ and maximum assortment capacity $K$) up to logarithmic factors. Our results close an $O(\sqrt{K})$ gap between upper and lower regret bounds from existing works.

stat.ML↗