arXiv subjects

Bo Chen

Publications and source records attributed to Bo Chen.

At least 397 records · Page 22Linked to original sources

arXiv2019

Low Power Inference for On-Device Visual Recognition with a Quantization-Friendly Solution

The IEEE Low-Power Image Recognition Challenge (LPIRC) is an annual competition started in 2015 that encourages joint hardware and software solutions for computer vision systems with low latency and power. Track 1 of the competition in 2018 focused on the innovation of software solutions with fixed inference engine and hardware. This decision allows participants to submit models online and not worry about building and bringing custom hardware on-site, which attracted a historically large number of submissions. Among the diverse solutions, the winning solution proposed a quantization-friendly framework for MobileNets that achieves an accuracy of 72.67% on the holdout dataset with an average latency of 27ms on a single CPU core of Google Pixel2 phone, which is superior to the best real-time MobileNet models at the time.

cs.CV↗

arXiv2019

MPC for Energy Efficient HVAC Control with Humidity and Latent Heat Considerations

Even though energy efficient climate control of buildings using model predictive control (MPC) has been widely investigated, most MPC formulations ignore humidity and latent heat. The inclusion of moisture makes the problem considerably more challenging, primarily since a cooling and dehumidifying coil model which accounts for both sensible and latent heat transfers is needed. In this work, we propose an MPC controller in which humidity and latent heat are incorporated in a principled manner. We construct low order data-driven models of a cooling and dehumidifying coil that can be used in the MPC formulation. The resulting controller's performance is tested in simulation using a plant that differs significantly from the model used by the optimizer. Additionally, the performance of the proposed controller is compared with that of a naive MPC controller which does not explicitly consider humidity, and also to that of a conventional rule-based controller. Simulations show that the proposed MPC controller outperforms the other two consistently. It is also observed that the naive MPC formulation which does not consider humidity leads to poor humidity control under certain conditions. Such violations in humidity can adversely affect occupant comfort and health.

eess.SY↗

arXiv2019

Sentence Rewriting for Semantic Parsing

A major challenge of semantic parsing is the vocabulary mismatch problem between natural language and target ontology. In this paper, we propose a sentence rewriting based semantic parsing method, which can effectively resolve the mismatch problem by rewriting a sentence into a new form which has the same structure with its target logical form. Specifically, we propose two sentence-rewriting methods for two common types of mismatch: a dictionary-based method for 1-N mismatch and a template-based method for N-1 mismatch. We evaluate our entence rewriting based semantic parser on the benchmark semantic parsing dataset -- WEBQUESTIONS. Experimental results show that our system outperforms the base system with a 3.4% gain in F1, and generates logical forms more accurately and parses sentences more robustly.

cs.CL↗

arXiv2019

Non-topological Origin of the Planar Hall Effect in Type-II Dirac Semimetal NiTe2

Dirac and Weyl semimetals are new discovered topological nontrivial materials with the linear band dispersions around the Dirac/Weyl points. When applying non-orthogonal electric current and magnetic field, an exotic phenomenon called chiral anomaly arises and negative longitudinal resistance can be detected. Recently, a new phenomenon named planer Hall effect (PHE) is considered to be another indication of chiral anomaly which has been observed in many topological semimetals. However, it still remains a question that is the PHE only attributed to chiral anomaly? Here we demonstrate the PHE in a new-discovered type-II Dirac semimetal NiTe2 by low temperature transport. However, after detailed analysis, we conclude that the PHE results from the trivial orbital magnetoresistance. This work reveals that PHE is not a sufficient condition of chiral anomaly and one need to take special care of other non-topological contribution in such studies.

cond-mat.mes-hall↗

arXiv2019

Deep Poisson gamma dynamical systems

We develop deep Poisson-gamma dynamical systems (DPGDS) to model sequentially observed multivariate count data, improving previously proposed models by not only mining deep hierarchical latent structure from the data, but also capturing both first-order and long-range temporal dependencies. Using sophisticated but simple-to-implement data augmentation techniques, we derived closed-form Gibbs sampling update equations by first backward and upward propagating auxiliary latent counts, and then forward and downward sampling latent variables. Moreover, we develop stochastic gradient MCMC inference that is scalable to very long multivariate count time series. Experiments on both synthetic and a variety of real-world data demonstrate that the proposed model not only has excellent predictive performance, but also provides highly interpretable multilayer latent structure to represent hierarchical and temporal information propagation.

stat.ML↗

arXiv2018

On the error analyses of polarization measurements of the white-light coronagraph aboard ASO-S

The Advanced Space-based Solar Observatory (ASO-S) mission aims to explore two most spectacular eruptions in the Sun: solar flares and coronal mass ejections (CMEs), and their magnetism. For the studies of CMEs, the payload Lyman-alpha Solar Telescope (LST) has been proposed. It includes a traditional white-light coronagraph and a Lyman-alpha coronagraph which opens a new window to CME observations. Polarization measurements taken by white-light coronagraphs are crucial to derive fundamental physical parameters of CMEs. To make such measurements, there are two options of Stokes polarimeter which have been used by existing white-light coronagraphs for space missions. One uses a single or triple linear polarizers, the other involves both a half-wave plate and a linear polarizer. We find that the former option subjects to less uncertainty in the derived Stokes vector propagated from detector noise. The latter option involves two plates which are prone to internal reflections and may have a reduced transmission factor. Therefore, the former option is adopted as our Stokes polarimeter scheme for LST. Based on the parameters of the intended linear polarizer(s) colorPol provided by CODIXX and the half-wave plate 2-APW-L2-012C by Altechna, it is further shown that the imperfect maximum transmittance of the polarizer significantly increases the variance amplification of Stokes vector by at least about 50% when compared with the ideal case. The relative errors of Stokes vector caused by the imperfection of colorPol polarizer and the uncertainty due to the polarizer assembling in the telescope are estimated to be about 5%. Among the considered parameters, we find that the dominant error comes from the uncertainty in the maximum transmittance of the polarizer.

astro-ph.IM↗

arXiv2018

Thermal-driven Flow inside Graphene Channels for Water Desalination

A novel concept of membrane process in thermal-driven system is proposed for water desalination. By means of molecular dynamics simulations, we show fast water transport through graphene galleries at a temperature gradient. Water molecules are driven to migrate through nanometer-wide graphene channels from cold reservoir to hot reservoir by the effect of thermal creep flow. Reducing the interlayer spacing to 6.5 Å, an abrupt escalation occurs in water permeation between angstrom-distance graphene slabs. The change from disordered bulklike water to quasi-square structure have been found under this extremely confined condition. This leads to a transition to subcontinuum transport. Water molecules perform collective diffusion behaviors inside graphene channels. The special transport processes with structure change convert thermal energy into motion without dissipation, resulting in unexpected high water permeability. The thermal-driven system reaches maximum flowrate at temperature variance of 80 K, corresponding to the quantity at pressure difference up to 10^5 bar in commercial reverse osmosis processes and 230 bar in pressure-driven slip flow. Our results also reveal the movement of saline ions influenced by thermophoretic effect, which complement the geometry limitation at greater layer spacing, enhancing the blockage of ions. This finding aims to provide an innovational idea of developing a high-efficiency desalination technology able to utilize various forms of energy.

cond-mat.soft↗

arXiv2018

On the Secrecy Unicast Throughput Performance of NOMA Assisted Multicast-Unicast Streaming With Partial Channel Information

This paper considers a downlink single-cell non-orthogonal multiple access (NOMA) network with uniformly deployed users, while a mixed multicast and unicast traffic scenario is taken into account. By guaranteeing the quality of service (QoS) of the multicast traffic, the multicast outage and secrecy unicast throughput performance is evaluated. In particular, two types of partial channel state information (CSI), namely the imperfect CSI and CSI based on second order statistics (SOS), are investigated in the analytical framework. For both two cases, the closed-form approximations for the multicast outage probability and secrecy unicast throughput are derived except that the approximation for the secrecy unicast throughput in SOS-based CSI only concerns two users. As to the multicast outage probability, the simulation results demonstrate that the NOMA scheme considered in both two cases achieves superior performance compared to the traditional orthogonal multiple access (OMA) scheme, while for the secrecy unicast throughput, the NOMA scheme shows great advantages over the OMA scheme in good condition (high signal-to-noise ratio (SNR)), but is inferior to the OMA scheme in poor condition (low SNR). Moreover, both two cases achieve similar performance except that the NOMA scheme with imperfect CSI obtains larger secrecy unicast throughput than that based on SOS under high SNR condition. Finally, the provided numerical results also confirm that the derived approximations of both two cases for the multicast outage probability and secrecy unicast throughput match well with the Monte Carlo simulations.

cs.IT↗

arXiv2018

2018 Low-Power Image Recognition Challenge

The Low-Power Image Recognition Challenge (LPIRC, https://rebootingcomputing.ieee.org/lpirc) is an annual competition started in 2015. The competition identifies the best technologies that can classify and detect objects in images efficiently (short execution time and low energy consumption) and accurately (high precision). Over the four years, the winners' scores have improved more than 24 times. As computer vision is widely used in many battery-powered systems (such as drones and mobile phones), the need for low-power computer vision will become increasingly important. This paper summarizes LPIRC 2018 by describing the three different tracks and the winners' solutions.

cs.CV↗

arXiv2018

NetAdapt: Platform-Aware Neural Network Adaptation for Mobile Applications

This work proposes an algorithm, called NetAdapt, that automatically adapts a pre-trained deep neural network to a mobile platform given a resource budget. While many existing algorithms simplify networks based on the number of MACs or weights, optimizing those indirect metrics may not necessarily reduce the direct metrics, such as latency and energy consumption. To solve this problem, NetAdapt incorporates direct metrics into its adaptation algorithm. These direct metrics are evaluated using empirical measurements, so that detailed knowledge of the platform and toolchain is not required. NetAdapt automatically and progressively simplifies a pre-trained network until the resource budget is met while maximizing the accuracy. Experiment results show that NetAdapt achieves better accuracy versus latency trade-offs on both mobile CPU and mobile GPU, compared with the state-of-the-art automated network simplification algorithms. For image classification on the ImageNet dataset, NetAdapt achieves up to a 1.7$\times$ speedup in measured inference latency with equal or higher accuracy on MobileNets (V1&V2).

cs.CV↗

arXiv2018

Analysis of the time domain acoustic scattering from open cavities

This paper is concerned with the analysis of the time domain acoustic scattering from two-dimensional open cavities. A transparent boundary condition is developed to reformulate the scattering problem into an equivalent initial boundary value problem in the interior domain of the open cavity. The well-posedness, that is, the existence, uniqueness and stability of the solution to the reduced problem are studied via a "Laplace domain" to time domain analysis. Moreover, time domain boundary integral equations for the reduced problem are established.

math.AP↗

arXiv2018

Sequence-to-Action: End-to-End Semantic Graph Generation for Semantic Parsing

This paper proposes a neural semantic parsing approach -- Sequence-to-Action, which models semantic parsing as an end-to-end semantic graph generation process. Our method simultaneously leverages the advantages from two recent promising directions of semantic parsing. Firstly, our model uses a semantic graph to represent the meaning of a sentence, which has a tight-coupling with knowledge bases. Secondly, by leveraging the powerful representation learning and prediction ability of neural network models, we propose a RNN model which can effectively map sentences to action sequences for semantic graph generation. Experiments show that our method achieves state-of-the-art performance on OVERNIGHT dataset and gets competitive performance on GEO and ATIS datasets.

cs.CL↗

arXiv2018

A New Approach to Linear/Nonlinear Distributed Fusion Estimation Problem

Disturbance noises are always bounded in a practical system, while fusion estimation is to best utilize multiple sensor data containing noises for the purpose of estimating a quantity--a parameter or process. However, few results are focused on the information fusion estimation problem under bounded noises. In this paper, we study the distributed fusion estimation problem for linear time-varying systems and nonlinear systems with bounded noises, where the addressed noises do not provide any statistical information, and are unknown but bounded. When considering linear time-varying fusion systems with bounded noises, a new local Kalman-like estimator is designed such that the square error of the estimator is bounded as time goes to $\infty$. A novel constructive method is proposed to find an upper bound of fusion estimation error, then a convex optimization problem on the design of an optimal weighting fusion criterion is established in terms of linear matrix inequalities, which can be solved by standard software packages. Furthermore, according to the design method of linear time-varying fusion systems, each local nonlinear estimator is derived for nonlinear systems with bounded noises by using Taylor series expansion, and a corresponding distributed fusion criterion is obtained by solving a convex optimization problem. Finally, target tracking system and localization of a mobile robot are given to show the advantages and effectiveness of the proposed methods.

eess.SY↗

arXiv2018

Analytical calculation of pressure for confined atomic and molecular systems using the eXtreme-Pressure Polarizable Continuum Model

We show that the pressure acting on atoms and molecular systems within the compression cavity of the eXtreme-Pressure Polarizable Continuum method can be expressed in terms of the electron density of the systems and of the Pauli-repulsion confining potential. The analytical expression holds for spherical cavities as well as for cavities constructed from van der Waals spheres of the constituting atoms of the molecular systems.

physics.chem-ph↗

arXiv2018

MorphNet: Fast & Simple Resource-Constrained Structure Learning of Deep Networks

We present MorphNet, an approach to automate the design of neural network structures. MorphNet iteratively shrinks and expands a network, shrinking via a resource-weighted sparsifying regularizer on activations and expanding via a uniform multiplicative factor on all layers. In contrast to previous approaches, our method is scalable to large networks, adaptable to specific resource constraints (e.g. the number of floating-point operations per inference), and capable of increasing the network's performance. When applied to standard network architectures on a wide variety of datasets, our approach discovers novel structures in each domain, obtaining higher performance while respecting the resource constraint.

cs.LG↗

arXiv2018

Oscillating planar Hall response from the surface electrons in bulk crystal Sn doped Bi1.1Sb0.9Te2S

We report the low-temperature magneto-transport in the bulk-insulating single crystal of topological insulator Sn doped Bi1.1Sb0.9Te2S. The Shubnikov-de Haas oscillations appear with their reciprocal frequency proportional to cos/theta , demonstrating the dominant transport of topological surface states. While the magnetic field is rotating in the sample surface, the planar Hall effect arises with sizeable oscillations following a relation of cos/theta sin/theta . Its amplitude reaches the maximum at the lowest temperature and drops to nearly zero at the temperature higher than 100 K. All these evidences consolidate such planar Hall oscillations as a new golden criterion on the topological surface transport.

cond-mat.mes-hall↗

arXiv2018

Simple networks on complex cellular automata: From de Bruijn diagrams to jump-graphs

We overview networks which characterise dynamics in cellular automata. These networks are derived from one-dimensional cellular automaton rules and global states of the automaton evolution: de Bruijn diagrams, subsystem diagrams, basins of attraction, and jump-graphs. These networks are used to understand properties of spatially-extended dynamical systems: emergence of non-trivial patterns, self-organisation, reversibility and chaos. Particular attention is paid to networks determined by travelling self-localisations, or gliders.

nlin.CG↗

arXiv2018

Tree-Structured Neural Machine for Linguistics-Aware Sentence Generation

Different from other sequential data, sentences in natural language are structured by linguistic grammars. Previous generative conversational models with chain-structured decoder ignore this structure in human language and might generate plausible responses with less satisfactory relevance and fluency. In this study, we aim to incorporate the results from linguistic analysis into the process of sentence generation for high-quality conversation generation. Specifically, we use a dependency parser to transform each response sentence into a dependency tree and construct a training corpus of sentence-tree pairs. A tree-structured decoder is developed to learn the mapping from a sentence to its tree, where different types of hidden states are used to depict the local dependencies from an internal tree node to its children. For training acceleration, we propose a tree canonicalization method, which transforms trees into equivalent ternary trees. Then, with a proposed tree-structured search method, the model is able to generate the most probable responses in the form of dependency trees, which are finally flattened into sequences as the system output. Experimental results demonstrate that the proposed X2Tree framework outperforms baseline methods over 11.15% increase of acceptance ratio.

cs.AI↗