arXiv subjects

Qiang Liu

Publications and source records attributed to Qiang Liu.

At least 433 records · Page 24Linked to original sources

arXiv2020

Black-box Off-policy Estimation for Infinite-Horizon Reinforcement Learning

Off-policy estimation for long-horizon problems is important in many real-life applications such as healthcare and robotics, where high-fidelity simulators may not be available and on-policy evaluation is expensive or impossible. Recently, \cite{liu18breaking} proposed an approach that avoids the \emph{curse of horizon} suffered by typical importance-sampling-based methods. While showing promising results, this approach is limited in practice as it requires data be drawn from the \emph{stationary distribution} of a \emph{known} behavior policy. In this work, we propose a novel approach that eliminates such limitations. In particular, we formulate the problem as solving for the fixed point of a certain operator. Using tools from Reproducing Kernel Hilbert Spaces (RKHSs), we develop a new estimator that computes importance ratios of stationary distributions, without knowledge of how the off-policy data are collected. We analyze its asymptotic consistency and finite-sample generalization. Experiments on benchmarks verify the effectiveness of our approach.

cs.LG↗

arXiv2020

Detection and Classification of Astronomical Targets with Deep Neural Networks in Wide Field Small Aperture Telescopes

Wide field small aperture telescopes are widely used for optical transient observations. Detection and classification of astronomical targets in observed images are the most important and basic step. In this paper, we propose an astronomical targets detection and classification framework based on deep neural networks. Our framework adopts the concept of the Faster R-CNN and uses a modified Resnet-50 as backbone network and a Feature Pyramid Network to extract features from images of different astronomical targets. To increase the generalization ability of our framework, we use both simulated and real observation images to train the neural network. After training, the neural network could detect and classify astronomical targets automatically. We test the performance of our framework with simulated data and find that our framework has almost the same detection ability as that of the traditional method for bright and isolated sources and our framework has 2 times better detection ability for dim targets, albeit all celestial objects detected by the traditional method can be classified correctly. We also use our framework to process real observation data and find that our framework can improve 25 % detection ability than that of the traditional method when the threshold of our framework is 0.6. Rapid discovery of transient targets is quite important and we further propose to install our framework in embedded devices such as the Nvidia Jetson Xavier to achieve real-time astronomical targets detection and classification abilities.

astro-ph.IM↗

arXiv2020

Stein Variational Inference for Discrete Distributions

Gradient-based approximate inference methods, such as Stein variational gradient descent (SVGD), provide simple and general-purpose inference engines for differentiable continuous distributions. However, existing forms of SVGD cannot be directly applied to discrete distributions. In this work, we fill this gap by proposing a simple yet general framework that transforms discrete distributions to equivalent piecewise continuous distributions, on which the gradient-free SVGD is applied to perform efficient approximate inference. The empirical results show that our method outperforms traditional algorithms such as Gibbs sampling and discontinuous Hamiltonian Monte Carlo on various challenging benchmarks of discrete graphical models. We demonstrate that our method provides a promising tool for learning ensembles of binarized neural network (BNN), outperforming other widely used ensemble methods on learning binarized AlexNet on CIFAR-10 dataset. In addition, such transform can be straightforwardly employed in gradient-free kernelized Stein discrepancy to perform goodness-of-fit (GOF) test on discrete distributions. Our proposed method outperforms existing GOF test methods for intractable discrete distributions.

cs.LG↗

arXiv2020

Statistical Adaptive Stochastic Gradient Methods

We propose a statistical adaptive procedure called SALSA for automatically scheduling the learning rate (step size) in stochastic gradient methods. SALSA first uses a smoothed stochastic line-search procedure to gradually increase the learning rate, then automatically switches to a statistical method to decrease the learning rate. The line search procedure ``warms up'' the optimization process, reducing the need for expensive trial and error in setting an initial learning rate. The method for decreasing the learning rate is based on a new statistical test for detecting stationarity when using a constant step size. Unlike in prior work, our test applies to a broad class of stochastic gradient algorithms without modification. The combined method is highly robust and autonomous, and it matches the performance of the best hand-tuned learning rate schedules in our experiments on several deep learning tasks.

stat.ML↗

arXiv2020

Two-dimensional-controlled high-order modes and vortex beams from an intracavity mode converter laser

We present a novel scheme of structured light laser with an astigmatic mode converter (AMC) as intracavity element, first enabling the generation of Hermite-Gaussian (HG) modes with fully controlled two-dimensional (2D) indices (m,n) and vortex beams carrying orbital angular momentum (OAM) directly from cavity. The 2D tunability was realized by controlling the off-axis displacements of both pump and intracavity AMC. The output HGm,n beam could be externally converted into OAM beam with 2D tunable radial and azimuthal indices (p,l). With the certain parameter control, vortex beam carrying OAM also could be directly generated from the cavity. Our setup provides a compact and concise structured light source. It has great potential in extending various applications of optical tweezers, communications, and nonlinearity.

physics.optics↗

arXiv2020

MaxUp: A Simple Way to Improve Generalization of Neural Network Training

We propose \emph{MaxUp}, an embarrassingly simple, highly effective technique for improving the generalization performance of machine learning models, especially deep neural networks. The idea is to generate a set of augmented data with some random perturbations or transforms and minimize the maximum, or worst case loss over the augmented data. By doing so, we implicitly introduce a smoothness or robustness regularization against the random perturbations, and hence improve the generation performance. For example, in the case of Gaussian perturbation, \emph{MaxUp} is asymptotically equivalent to using the gradient norm of the loss as a penalty to encourage smoothness. We test \emph{MaxUp} on a range of tasks, including image classification, language modeling, and adversarial certification, on which \emph{MaxUp} consistently outperforms the existing best baseline methods, without introducing substantial computational overhead. In particular, we improve ImageNet classification from the state-of-the-art top-1 accuracy $85.5\%$ without extra data to $85.8\%$. Code will be released soon.

cs.LG↗

arXiv2020

Accelerating Metropolis-within-Gibbs sampler with localized computations of differential equations

Inverse problem is ubiquitous in science and engineering, and Bayesian methodologies are often used to infer the underlying parameters. For high dimensional temporal-spatial models, classical Markov chain Monte Carlo (MCMC) methods are often slow to converge, and it is necessary to apply Metropolis-within-Gibbs (MwG) sampling on parameter blocks. However, the computation cost of each MwG iteration is typically $O(n^2)$, where $n$ is the model dimension. This can be too expensive in practice. This paper introduces a new reduced computation method to bring down the computation cost to $O(n)$, for the inverse initial value problem of a stochastic differential equation (SDE) with local interactions. The key observation is that each MwG proposal is only different from the original iterate at one parameter block, and this difference will only propagate within a local domain in the SDE computations. Therefore we can approximate the global SDE computation with a surrogate updated only within the local domain for reduced computation cost. Both theoretically and numerically, we show that the approximation errors can be controlled by the local domain size. We discuss how to implement the local computation scheme using Euler--Maruyama and 4th order Runge--Kutta methods. We numerically demonstrate the performance of the proposed method with the Lorenz 96 model and a linear stochastic flow model.

stat.CO↗

arXiv2020

A Kernel Loss for Solving the Bellman Equation

Value function learning plays a central role in many state-of-the-art reinforcement-learning algorithms. Many popular algorithms like Q-learning do not optimize any objective function, but are fixed-point iterations of some variant of Bellman operator that is not necessarily a contraction. As a result, they may easily lose convergence guarantees, as can be observed in practice. In this paper, we propose a novel loss function, which can be optimized using standard gradient-based methods without risking divergence. The key advantage is that its gradient can be easily approximated using sampled transitions, avoiding the need for double samples required by prior algorithms like residual gradient. Our approach may be combined with general function classes such as neural networks, on either on- or off-policy data, and is shown to work reliably and effectively in several benchmarks.

cs.LG↗

arXiv2020

Attributed Multi-Relational Attention Network for Fact-checking URL Recommendation

To combat fake news, researchers mostly focused on detecting fake news and journalists built and maintained fact-checking sites (e.g., Snopes.com and Politifact.com). However, fake news dissemination has been greatly promoted via social media sites, and these fact-checking sites have not been fully utilized. To overcome these problems and complement existing methods against fake news, in this paper we propose a deep-learning based fact-checking URL recommender system to mitigate impact of fake news in social media sites such as Twitter and Facebook. In particular, our proposed framework consists of a multi-relational attentive module and a heterogeneous graph attention network to learn complex/semantic relationship between user-URL pairs, user-user pairs, and URL-URL pairs. Extensive experiments on a real-world dataset show that our proposed framework outperforms eight state-of-the-art recommendation models, achieving at least 3~5.3% improvement.

cs.IR↗

arXiv2019

Exploration via Hindsight Goal Generation

Goal-oriented reinforcement learning has recently been a practical framework for robotic manipulation tasks, in which an agent is required to reach a certain goal defined by a function on the state space. However, the sparsity of such reward definition makes traditional reinforcement learning algorithms very inefficient. Hindsight Experience Replay (HER), a recent advance, has greatly improved sample efficiency and practical applicability for such problems. It exploits previous replays by constructing imaginary goals in a simple heuristic way, acting like an implicit curriculum to alleviate the challenge of sparse reward signal. In this paper, we introduce Hindsight Goal Generation (HGG), a novel algorithmic framework that generates valuable hindsight goals which are easy for an agent to achieve in the short term and are also potential for guiding the agent to reach the actual goal in the long term. We have extensively evaluated our goal generation algorithm on a number of robotic manipulation tasks and demonstrated substantially improvement over the original HER in terms of sample efficiency.

cs.LG↗

arXiv2019

Comparison of counterstreaming suprathermal electron signatures of ICMEs with and without magnetic cloud: are all ICMEs flux ropes?

Magnetic clouds (MCs), as large-scale interplanetary magnetic flux ropes, are usually still connected to the sun at both ends near 1 AU. Many researchers believe that all non-MC interplanetary coronal mass ejections (ICMEs) also have magnetic flux rope structures, which are inconspicuous because the observing spacecraft crosses the flanks of the rope structures. If so, the field lines of non-MC ICMEs should also be usually connected to the Sun on both ends. Then we want to know whether the field lines of most non-MC ICMEs are still connected to the sun at both ends or not. This study examined the counterstreaming suprathermal electron (CSE) signatures of 266 ICMEs observed by the \emph{Advanced Composition Explorer} (\emph{ACE}) spacecraft from 1998 to 2008 and compared the CSE signatures of MCs and non-MC ICMEs. Results show that only 10 of the 101 MC events ($9.9\%$ ) and 75 of the 171 non-MC events ($43.9\%$) have no CSEs. Moreover, 21 of the non-MC ICMEs have high CSE percentages (more than $70\%$) and show relatively stable magnetic field components with slight rotations, which are in line with the expectations that spacecraft passes through the flank of magnetic flux ropes. So the 21 events may be magnetic flux ropes but the \emph{ACE} spacecraft passes through their flanks of magnetic flux ropes. Considering that most other non-MC events have disordered magnetic fields, we suggest that some non-MC ICMEs inherently have disordered magnetic fields, namely have no magnetic flux rope structures.

physics.space-ph↗

arXiv2019

Learning Preferences and Demands in Visual Recommendation

Visual information is an important factor in recommender systems, in which users' selections consist of two components: \emph{preferences} and \emph{demands}. Some studies has been done for modeling users' preferences in visual recommendation. However, conventional methods models items in a common visual feature space, which may fail in capturing \emph{styles} of items. We propose a DeepStyle method for learning style features of items. DeepStyle eliminates the categorical information of items, which is dominant in the original visual feature space, based on a Convolutional Neural Networks (CNN) architecture. For modeling users' demands on different categories of items, the problem can be formulated as recommendation with contextual and sequential information. To solve this problem, we propose a Context-Aware Gated Recurrent Unit (CA-GRU) method, which can capture sequential and contextual information simultaneously. Furthermore, the aggregation of prediction on preferences and demands, i.e., prediction generated by DeepStyle and CA-GRU, can model users' selection behaviors more completely. Experiments conducted on real-world datasets illustrates the effectiveness of our proposed methods in visual recommendation.

cs.IR↗

arXiv2019

Stein Variational Gradient Descent With Matrix-Valued Kernels

Stein variational gradient descent (SVGD) is a particle-based inference algorithm that leverages gradient information for efficient approximate inference. In this work, we enhance SVGD by leveraging preconditioning matrices, such as the Hessian and Fisher information matrix, to incorporate geometric information into SVGD updates. We achieve this by presenting a generalization of SVGD that replaces the scalar-valued kernels in vanilla SVGD with more general matrix-valued kernels. This yields a significant extension of SVGD, and more importantly, allows us to flexibly incorporate various preconditioning matrices to accelerate the exploration in the probability landscape. Empirical results show that our method outperforms vanilla SVGD and a variety of baseline approaches over a range of real-world Bayesian inference tasks.

stat.ML↗

arXiv2019

Splitting Steepest Descent for Growing Neural Architectures

We develop a progressive training approach for neural networks which adaptively grows the network structure by splitting existing neurons to multiple off-springs. By leveraging a functional steepest descent idea, we derive a simple criterion for deciding the best subset of neurons to split and a splitting gradient for optimally updating the off-springs. Theoretically, our splitting strategy is a second-order functional steepest descent for escaping saddle points in an $\infty$-Wasserstein metric space, on which the standard parametric gradient descent is a first-order steepest descent. Our method provides a new computationally efficient approach for optimizing neural network structures, especially for learning lightweight neural architectures in resource-constrained settings.

cs.LG↗

arXiv2019

Learning-Assisted Secure End-to-End Network Slicing for Cyber-Physical Systems

There is a pressing need to interconnect physical systems such as power grid and vehicles for efficient management and safe operations. Owing to the diverse features of physical systems, there is hardly a one-size-fits-all networking solution for developing cyber-physical systems. Network slicing is a promising technology that allows network operators to create multiple virtual networks on top of a shared network infrastructure. These virtual networks can be tailored to meet the requirements of different cyber-physical systems. However, it is challenging to design secure network slicing solutions that can efficiently create end-to-end network slices for diverse cyber-physical systems. In this article, we discuss the challenges and security issues of network slicing, study learning-assisted network slicing solutions, and analyze their performance under the denial-of-service attack. We also present a design and implementation of a small-scale testbed for evaluating the network slicing solutions.

cs.NI↗

arXiv2019

A General Framework for Empirical Bayes Estimation in Discrete Linear Exponential Family

We develop a Nonparametric Empirical Bayes (NEB) framework for compound estimation in the discrete linear exponential family, which includes a wide class of discrete distributions frequently arising from modern big data applications. We propose to directly estimate the Bayes shrinkage factor in the generalized Robbins' formula via solving a scalable convex program, which is carefully developed based on a RKHS representation of the Stein's discrepancy measure. The new NEB estimation framework is flexible for incorporating various structural constraints into the data driven rule, and provides a unified approach to compound estimation with both regular and scaled squared error losses. We develop theory to show that the class of NEB estimators enjoys strong asymptotic properties. Comprehensive simulation studies as well as analyses of real data examples are carried out to demonstrate the superiority of the NEB estimator over competing methods.

math.ST↗

arXiv2019

Doubly Robust Bias Reduction in Infinite Horizon Off-Policy Estimation

Infinite horizon off-policy policy evaluation is a highly challenging task due to the excessively large variance of typical importance sampling (IS) estimators. Recently, Liu et al. (2018a) proposed an approach that significantly reduces the variance of infinite-horizon off-policy evaluation by estimating the stationary density ratio, but at the cost of introducing potentially high biases due to the error in density ratio estimation. In this paper, we develop a bias-reduced augmentation of their method, which can take advantage of a learned value function to obtain higher accuracy. Our method is doubly robust in that the bias vanishes when either the density ratio or the value function estimation is perfect. In general, when either of them is accurate, the bias can also be reduced. Both theoretical and empirical results show that our method yields significant advantages over previous methods.

cs.LG↗

arXiv2019

Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm

We propose a general purpose variational inference algorithm that forms a natural counterpart of gradient descent for optimization. Our method iteratively transports a set of particles to match the target distribution, by applying a form of functional gradient descent that minimizes the KL divergence. Empirical studies are performed on various real world models and datasets, on which our method is competitive with existing state-of-the-art methods. The derivation of our method is based on a new theoretical result that connects the derivative of KL divergence under smooth transforms with Stein's identity and a recently proposed kernelized Stein discrepancy, which is of independent interest.

stat.ML↗