Still searching Arxiv, refresh for possibly better results.
Results for "Liwei Wang"
total 39709took 0.13s
The Expressive Power of Neural Networks: A View from the WidthSep 08 2017Nov 01 2017The expressive power of neural networks is important for understanding deep learning. Most existing works consider this problem from the view of the depth of a network. In this paper, we study how width affects the expressiveness of neural networks. Classical ... More Learning Region Features for Object DetectionMar 19 2018While most steps in the modern object detection methods are learnable, the region feature extraction step remains largely hand-crafted, featured by RoI pooling methods. This work proposes a general viewpoint that unifies existing region feature extraction ... More Training Deeper Convolutional Networks with Deep SupervisionMay 11 2015One of the most promising ways of improving the performance of deep convolutional neural networks is by increasing the number of convolutional layers. However, adding layers makes training more difficult and computationally expensive. In order to train ... More Gradient Descent Finds Global Minima of Deep Neural NetworksNov 09 2018Feb 04 2019Gradient descent finds a global minimum in training deep neural networks despite the objective function being non-convex. The current paper proves gradient descent achieves zero training loss in polynomial time for a deep over-parameterized neural network ... More Pairwise Constraint Propagation on Multi-View DataJan 18 2015This paper presents a graph-based learning approach to pairwise constraint propagation on multi-view data. Although pairwise constraint propagation has been studied extensively, pairwise constraints are usually defined over pairs of data points from a ... More KBGAN: Adversarial Learning for Knowledge Graph EmbeddingsNov 11 2017Apr 16 2018We introduce KBGAN, an adversarial learning framework to improve the performances of a wide range of existing knowledge graph embedding models. Because knowledge graphs typically only contain positive facts, sampling useful negative training examples ... More Broken Dynamic Symmetry and Phase Transition PrecursorFeb 22 2013Symmetry breaking is a central concept of Landau phase transition theory, which, however, only considers time-averaged static symmetry of crystal lattice while neglects dynamic symmetry of lattice vibrations thus fails to explain the ubiquitous transformation ... More Learning Deep Structure-Preserving Image-Text EmbeddingsNov 19 2015Apr 14 2016This paper proposes a method for learning joint embeddings of images and text using a two-branch neural network with multiple layers of linear projections followed by nonlinearities. The network is trained using a large margin objective that combines ... More Study on Timing Performance of a Readout Circuit for SiPMJun 07 2018In recent years, SiPM photoelectric devices have drawn much attention in the domain of time-of-flight-based positron emission tomography (TOF-PET). Using them to construct PET detectors with excellent coincidence time resolution (CTR) is always one of ... More Improved Silbey-Harris polaron ansatz for the spin-boson modelMar 27 2018In this paper, the well-known Silbey-Harris (SH) polaron ansatz for the spin-boson model is improved by adding orthogonal displaced Fock states. The obtained results for the ground state in all baths converge very quickly within finite displaced Fock ... More Efficient Private ERM for Smooth ObjectivesMar 29 2017May 24 2017In this paper, we consider efficient differentially private empirical risk minimization from the viewpoint of optimization algorithms. For strongly convex and smooth objectives, we prove that gradient descent with output perturbation not only achieves ... More A Theoretical Analysis of NDCG Type Ranking MeasuresApr 24 2013A central problem in ranking is to design a ranking measure for evaluation of ranking functions. In this paper we study, from a theoretical perspective, the widely used Normalized Discounted Cumulative Gain (NDCG)-type ranking measures. Although there ... More Weighted Bergman Projection on the Hartogs TriangleOct 22 2014Apr 26 2015We prove the $L^p$ regularity of the weighted Bergman projection on the Hartogs triangle, where the weights are powers of the distance to the singularity at the boundary. The restricted range of $p$ is proved to be sharp. By using a two-weight inequality ... More MedSTS: A Resource for Clinical Semantic Textual SimilarityAug 28 2018The wide adoption of electronic health records (EHRs) has enabled a wide range of applications leveraging EHR data. However, the meaningful use of EHR data largely depends on our ability to efficiently extract and consolidate information embedded in clinical ... More Learning to Navigate for Fine-grained ClassificationSep 02 2018Fine-grained classification is challenging due to the difficulty of finding discriminative features. Finding those subtle traits that fully characterize the object is not straightforward. To handle this circumstance, we propose a novel self-supervision ... More On the Depth of Deep Neural Networks: A Theoretical ViewJun 17 2015Nov 28 2015People believe that depth plays an important role in success of deep neural networks (DNN). However, this belief lacks solid theoretical justifications as far as we know. We investigate role of depth from perspective of margin bound. In margin bound, ... More Dual Learning for Machine TranslationNov 01 2016While neural machine translation (NMT) is making good progress in the past two years, tens of millions of bilingual sentence pairs are needed for its training. However, human labeling is very costly. To tackle this training data bottleneck, we develop ... More On Low Rank Approximation of Binary MatricesNov 05 2015We consider the problem of low rank approximation of binary matrices. Here we are given a $d \times n$ binary matrix $A$ and a small integer $k < d$. The goal is to find two binary matrices $U$ and $V$ of sizes $d \times k$ and $k \times n$ respectively, ... More Towards Binary-Valued Gates for Robust LSTM TrainingJun 08 2018Long Short-Term Memory (LSTM) is one of the most widely used recurrent structures in sequence modeling. It aims to use gates to control information flow (e.g., whether to skip some information or not) in the recurrent computations, although its practical ... More FRAGE: Frequency-Agnostic Word RepresentationSep 18 2018Continuous word representation (aka word embedding) is a basic building block in many neural network-based models used in natural language processing tasks. Although it is widely accepted that words with similar semantics should be close to each other ... More Solutions to the quantum Rabi model with two equivalent qubitsJan 25 2014May 29 2014Using extended coherent states, an analytically exact study has been carried out for the quantum Rabi model with two equivalent qubits. Compact transcendental functions of one variable have been derived leading to exact solutions. The energy spectrum ... More Randomness in Deconvolutional Networks for Visual RepresentationApr 02 2017Feb 20 2018Toward a deeper understanding on the inner work of deep neural networks, we investigate CNN (convolutional neural network) using DCN (deconvolutional network) and randomization technique, and gain new insights for the intrinsic property of this network ... More Invariant algebraic surfaces of the FitzHugh-Nagumo systemJan 04 2017In this paper, we characterize all the irreducible Darboux polynomials and polynomial first integrals of FitzHugh-Nagumo (F-N) system. The method of the weight homogeneous polynomials and the characteristic curves is widely used to give a complete classification ... More On the translates of general dyadic systems on $\R$Sep 04 2018Many techniques in harmonic analysis use the fact that a continuous object can be written as a sum (or an intersection) of dyadic counterparts, as long as those counterparts belong to a distinct dyadic system. Here we generalize the notion of distinct ... More Inaudible Voice CommandsAug 24 2017Voice assistants like Siri enable us to control IoT devices conveniently with voice commands, however, they also provide new attack opportunities for adversaries. Previous papers attack voice assistants with obfuscated voice commands by leveraging the ... More Smoothing Properties of the Friedrichs Operator on $L^p$ spacesSep 01 2017Dec 16 2017We show that the Friedrichs operator exhibits smoothing properties in the $L^p$ scale. In particular we prove that on any smoothly bounded pseudoconvex domain the Friedrichs operator maps $A^2(\Omega)$ to $A^p(\Omega)$ for some $p>2$. Solutions to the anisotropic quantum Rabi modelFeb 02 2015In this work, the anisotropic quantum Rabi model with different coupling strengths of the rotating-wave and counter-rotating wave terms is studied by using two kinds of extended coherent states (ECS). By the first kind of ECS, we can derive a so-called ... More A Bounded Formulation for The School Bus Scheduling ProblemMar 24 2018This paper proposes a new formulation for the school bus scheduling problem (SBSP) which optimizes starting times for schools and associated bus routes to minimize transportation cost. Specifically, the problem determines the minimum number of buses required ... More SQL-Rank: A Listwise Approach to Collaborative RankingFeb 28 2018Feb 06 2019In this paper, we propose a listwise approach for constructing user-specific rankings in recommendation systems in a collaborative fashion. We contrast the listwise approach to previous pointwise and pairwise approaches, which are based on treating either ... More Let the Cloud Watch Over Your IoT File SystemsFeb 17 2019Smart devices produce security-sensitive data and keep them in on-device storage for persistence. The current storage stack on smart devices, however, offers weak security guarantees: not only because the stack depends on a vulnerable commodity OS, but ... More