arXiv subjects

Ning Liu

Publications and source records attributed to Ning Liu.

At least 199 records · Page 11Linked to original sources

arXiv2020

Mono-b events from single stop production at the HL-LHC and HE-LHC

Top-squarks (stop) play an important role in SUSY naturalness. The stop pair production is considered as the most effective way to search for stop at the LHC. However, the collider signature of stop pair production is usually characterized by $t\bar{t}$ plus missing transverse energy, which is also predicted in many other non-supersymmetric models. On the other hand, the single stop production via the electroweak interaction can provide some distinctive signatures, and thus will help to confirm the existence of the stop. In this paper, we investigate the observability of the mono-$b$ events from the single stop production process $pp \to \tilde t_1 \tildeχ^-_1 \to b+ E\!\!\!\!/_T$ in a simplified MSSM framework where the higgsinos and stops are the only sparticles at the HL-LHC and HE-LHC. We find that the stop mass and the higgsino mass may be probed up to about 1.6 TeV and 550 GeV at $5σ$ level at the HE-LHC with the integrated luminosity ${\cal L} = 15~\text{ab}^{-1}$. We also present the $2σ$ exclusion limits of the stop mass at the HL-LHC and HE-LHC.

hep-ph↗

arXiv2020

Top quark as a probe of heavy Majorana neutrino at the LHC and future collider

Right-handed (RH) Majorana neutrinos play a crucial role in understanding the origin of neutrino mass, the nature of dark matter and the mechanism of matter-antimatter asymmetry. In this work, we investigate the observability of heavy Majorana neutrino through the top quark neutrinoless double beta decay process $t \to b \ell^+ \ell^+ j j$ at hadron colliders. By performing detector level simulation, we demonstrate that our method can give stronger limits on the light-heavy neutrino mixing parameters $|V_{eN, μN}|$ in the mass range of 15 GeV $< m_N <$ 80 GeV than other existing collider bounds.

hep-ph↗

arXiv2020

Universal Physical Camouflage Attacks on Object Detectors

In this paper, we study physical adversarial attacks on object detectors in the wild. Previous works mostly craft instance-dependent perturbations only for rigid or planar objects. To this end, we propose to learn an adversarial pattern to effectively attack all instances belonging to the same object category, referred to as Universal Physical Camouflage Attack (UPC). Concretely, UPC crafts camouflage by jointly fooling the region proposal network, as well as misleading the classifier and the regressor to output errors. In order to make UPC effective for non-rigid or non-planar objects, we introduce a set of transformations for mimicking deformable properties. We additionally impose optimization constraint to make generated patterns look natural to human observers. To fairly evaluate the effectiveness of different physical-world attacks, we present the first standardized virtual database, AttackScenes, which simulates the real 3D world in a controllable and reproducible environment. Extensive experiments suggest the superiority of our proposed UPC compared with existing physical adversarial attackers not only in virtual environments (AttackScenes), but also in real-world physical environments. Code and dataset are available at https://mesunhlf.github.io/index_physical.html.

cs.CV↗

arXiv2020

Noble-Metal-Free Photocatalytic Hydrogen Evolution Activity: The Impact of Ball Milling Anatase Nanopowders with TiH2

In this work, we demonstrate that a well-established and facile ball milling approach using mixtures of commercial anatase nanoparticles and TiH2 introduces noble-metal-free photocatalytic H2 activity to titania. We characterize this synergistic effect in view of the nature of defects, state of hydroxylation, and investigate the effect on the energetics and kinetics of electronic states and the resulting H2 evolution performance.

physics.chem-ph↗

arXiv2020

Large diameter TiO$_2$ nanotubes enable integration of conformed hierarchical and blocking layers for enhanced dye-sensitized solar cell efficiency

In the present work we grow anodic TiO$_2$ nanotube layer with tube diameter ~ 500 nm and an open tube mouth. We use this morphology in dye-sensitized solar cells (DSSCs) and show that these tubes allow the construction of hybrid hierarchical photoanode structures of nanotubes with a defined and wall-conformance TiO2 nanoparticles decoration. At the same time, the large diameter allows the successful establishment of an additional (insulating) blocking layer of SiO$_2$ or Al$_2$O$_3$. We show that this combination of hierarchical structure and blocking layer significantly enhances the solar cell efficiency by suppressing recombination reactions. In such a DSSC structure, the solar cell efficiency under back side illumination with AM1.5 illumination is enhanced from 5% neat tube to 7 %.

physics.app-ph↗

arXiv2020

Black Magic in Gray Titania: Noble-Metal-Free Photocatalytic H2 Evolution from Hydrogenated Anatase

"Black" TiO2 has gained increasing interest because of its outstanding properties and promising applications in a wide range of fields. Among the outstanding features of the material is that certain synthesis processes lead to the formation of an intrinsic co-catalytic center and thus enable noble-metal free photocatalytic H2 generation. In this work, we report "grey TiO2" by an appropriate hydrogenation treatment exhibits excellent photocatalytic hydrogen. In this case, by the employment of thermally stable and high-surface-area TiO2 nanoparticles as well as mesoporous particles as the hydrogenation precursor, the appropriate extent of reduction of TiO2 (coloration) and the formation of Ti3+ is the key for the efficient noble-metal-free photocatalytic H2 generation. The EPR results reveal that "grey TiO2" shows stronger Ti3+ feature at g ca. 1.93 than "black TiO2" contributing to the intrinsic catalytic center for H2 evolution.

physics.chem-ph↗

arXiv2020

Intrinsically Activated SrTiO3: Photocatalytic H2 Evolution from Neutral Aqueous Methanol Solution in the Absence of Any Noble Metal Cocatalyst

Noble metal cocatalysts are conventionally a crucial factor in oxide-semiconductor-based photocatalytic hydrogen generation. In the present work, we show that optimized high-temperature hydrogenation of commercially available strontium titanate (SrTiO3) powder can be used to engineer an intrinsic cocatalytic shell around nanoparticles that can create a photocatalyst that is highly effective without the use of any additional cocatalyst for hydrogen generation from neutral aqueous methanol solutions. This intrinsic activation effect can also be observed for SrTiO3[100] single crystal as well as Nb-doped SrTiO3 (100) single crystal. For all types of SrTiO3 samples (nanopowders and either of the single crystals), hydrogenation under optimum conditions leads to a surface-hydroxylated layer together with lattice defects visible by transmission electron microscopy, electron paramagnetic resonance (EPR), and photoluminescence (PL). Active samples provide states in a defective matrix -- this is in contrast to the inactive defects formed in other reductive atmospheres. In aqueous media, active SrTiO3 samples show a significant negative shift of the flatband potential (in photoelectrochemical as well as in capacitance data) and a lower charge-transfer resistance for photoexcited electrons. We therefore ascribe the remarkable cocatalyst-free activation of the material to a synergy between thermodynamics (altered interface energetics induced by hydroxylation) and kinetics (charge transfer mediation by suitable Ti3+ states).

cond-mat.mtrl-sci↗

arXiv2020

Optimized FTO seeding enables the growth of high efficient Ta-doped TiO$_2$ nanorod photoanodes

Tantalum doped rutile nanorods were hydrothermally grown on FTO substrates using a new seeding approach. This approach allows the incorporation of high concentrations of up to 4.8 at% tantalum as active doping and results in a significant enhancement of photoelectrochemical water splitting rate (1.8 mA/cm2 at a potential of +1.5 V vs RHE) which corresponds to ca. 1% photocurrent conversion efficiency under AM 1.5, 100 mW/cm2 simulated sunlight irradiation.

physics.chem-ph↗

arXiv2020

Black and White Anatase, Rutile and Mixed Forms: Band-Edges and Photocatalytic Activity

Here we investigate the band-level energetics of "black" hydrogenated titania in different polymorphs using in-situ photoelectrochemical measurements and XPS valence band measurements. We find that the conduction band of black rutile is higher in energy than in black anatase by 0.4 eV. For photocatalytic hydrogen generation, in a polymorph hetero-junction such as in black Degussa P25, thus black rutile can act as a photosensitizer while black anatase provides charge-mediation catalysis onto H2O to generate H2. By optimizing the thermal reduction conditions of black anatase/rutile junctions the H2 production can be significantly increased.

physics.chem-ph↗

arXiv2020

Noble metal free photocatalytic H$_2$ generation on black TiO$_2$: On the influence of crystal facets vs. crystal damage

In this study, we investigate noble metal free photocatalytic water splitting on natural anatase single crystal facets and on wafer slices of the [001] plane before and after these surfaces have been modified by high pressure hydrogenation (HPH) and hydrogen ion-implantation. We find that on the natural, intact low index planes photocatalytic H$_2$ evolution (in absence of noble metal co-catalyst) can only be achieved when the hydrogenation treatment is accompanied by the introduction of crystal damage, such as simple scratching, miscut in the wafer or by implantation damage. X-ray reflectivity (XRR), Raman, and optical reflection measurements show that plain hydrogenation leads to a ~ 1 nm thick black titania surface layer without activity, while a colorless, density modified and ~ 7 nm thick layer with broken crystal symmetry is present in the ion implanted surface. These results demonstrate that i) the H-treatment of an intact anatase surface needs to be combined with defect formation for catalytic activation, and ii) activation does not necessarily coincide with the presence of black color.

physics.app-ph↗

arXiv2020

Photocatalysis with TiO2 Nanotubes: Colorful Reactivity and Designing Site-Specific Photocatalytic Centers into TiO2 Nanotubes

Photocatalytic reactions on TiO2 have recently gained an enormous resurgence because of various new strategies and findings that promise to drastically increase efficiency and specificity of such reactions by modifications of the titania scaffold and chemistry. In view of geometry, in particular, anodic TiO2 nanotubes have attracted wide interest, as they allow a high degree of control over the separation of photogenerated charge carriers not only in photocatalytic reactions but also in photoelectrochemical reactions. A key advantage of ordered nanotube arrays is that nanotube modifications can be embedded site specifically into the tube wall; that is, cocatalysts, doping species, or junctions can be placed at highly defined desired locations (or with a desired regular geometry or pattern) along the tube wall. This allows an unprecedented level of engineering of energetics of reaction sites for catalytic and photocatalytic reactions, which target not only higher efficiencies but also the selectivity of reactions. Many recent tube alterations are of a morphologic nature (mesoporous structures, designed faceted crystallites, hybrids, or 1D structures), but a number of color-coded (namely, black, blue, red, green, gray) modifications have attracted wide interest because of the extension of the light absorption spectrum of titania in the visible range and because unique catalytic activity can be induced. The present Perspective gives an overview of TiO2 nanotubes in photocatalysis with an emphasis on the most recent advances in the use of nanotube arrays and discusses the underlying concepts in tailoring their photocatalytic reactivity.

physics.app-ph↗

arXiv2020

Interactive Human-Machine Learning Framework for Modelling of Ferroelectric-Dielectric Composites

Data driven materials discovery and optimization requires databases that are error free and experimentally verified. Performing material measurements are time-consuming and often restricted by the fact that material sample preparations are non-trivial, labour-intensive and expensive. Numerical modelling of materials has been studied over the years in order to address these issues and nowadays it has been developed at multi-scale and multi-physics levels. However, numerical models for nano-composites, especially for ferroelectrics are limited due to multiple unknowns including oxygen vacancy densities, grain sizes and domain boundaries existing in the system. In this work, we introduce a human-machine interactive learning framework by developing a scalable semi-empirical model to accurately predict material properties enabled by deep learning (DL). MgO-doped BST (BaxSr1-xTiO3) is selected as an example ferroelectric-dielectric composite for validation. The DL model transfer-learns the experimental features of materials from a measurement database which includes data for over 100 different ferroelectric composites collected by screening the published data and combining our own measurement data. The trained DL model is utilized in providing feedback to human researchers, who then refine computer model parameters accordingly, hence completing the interactive learning cycle. Finally, the developed DL model is applied to predict and optimise new ferroelectric-dielectric composites with the highest figure of merit (FOM) value.

cond-mat.mtrl-sci↗

arXiv2020

Transparent Classification with Multilayer Logical Perceptrons and Random Binarization

Models with transparent inner structure and high classification performance are required to reduce potential risk and provide trust for users in domains like health care, finance, security, etc. However, existing models are hard to simultaneously satisfy the above two properties. In this paper, we propose a new hierarchical rule-based model for classification tasks, named Concept Rule Sets (CRS), which has both a strong expressive ability and a transparent inner structure. To address the challenge of efficiently learning the non-differentiable CRS model, we propose a novel neural network architecture, Multilayer Logical Perceptron (MLLP), which is a continuous version of CRS. Using MLLP and the Random Binarization (RB) method we proposed, we can search the discrete solution of CRS in continuous space using gradient descent and ensure the discrete CRS acts almost the same as the corresponding continuous MLLP. Experiments on 12 public data sets show that CRS outperforms the state-of-the-art approaches and the complexity of the learned CRS is close to the simple decision tree. Source code is available at https://github.com/12wang3/mllp.

cs.LG↗

arXiv2020

Danger-aware Adaptive Composition of DRL Agents for Self-navigation

Self-navigation, referred as the capability of automatically reaching the goal while avoiding collisions with obstacles, is a fundamental skill required for mobile robots. Recently, deep reinforcement learning (DRL) has shown great potential in the development of robot navigation algorithms. However, it is still difficult to train the robot to learn goal-reaching and obstacle-avoidance skills simultaneously. On the other hand, although many DRL-based obstacle-avoidance algorithms are proposed, few of them are reused for more complex navigation tasks. In this paper, a novel danger-aware adaptive composition (DAAC) framework is proposed to combine two individually DRL-trained agents, obstacle-avoidance and goal-reaching, to construct a navigation agent without any redesigning and retraining. The key to this adaptive composition approach is that the value function outputted by the obstacle-avoidance agent serves as an indicator for evaluating the risk level of the current situation, which in turn determines the contribution of these two agents for the next move. Simulation and real-world testing results show that the composed Navigation network can control the robot to accomplish difficult navigation tasks, e.g., reaching a series of successive goals in an unknown and complex environment safely and quickly.

cs.RO↗

arXiv2019

Deep Compressed Pneumonia Detection for Low-Power Embedded Devices

Deep neural networks (DNNs) have been expanded into medical fields and triggered the revolution of some medical applications by extracting complex features and achieving high accuracy and performance, etc. On the contrast, the large-scale network brings high requirements of both memory storage and computation resource, especially for portable medical devices and other embedded systems. In this work, we first train a DNN for pneumonia detection using the dataset provided by RSNA Pneumonia Detection Challenge. To overcome hardware limitation for implementing large-scale networks, we develop a systematic structured weight pruning method with filter sparsity, column sparsity and combined sparsity. Experiments show that we can achieve up to 36x compression ratio compared to the original model with 106 layers, while maintaining no accuracy degradation. We evaluate the proposed methods on an embedded low-power device, Jetson TX2, and achieve low power usage and high energy efficiency.

eess.IV↗

arXiv2019

REQ-YOLO: A Resource-Aware, Efficient Quantization Framework for Object Detection on FPGAs

Deep neural networks (DNNs), as the basis of object detection, will play a key role in the development of future autonomous systems with full autonomy. The autonomous systems have special requirements of real-time, energy-efficient implementations of DNNs on a power-constrained system. Two research thrusts are dedicated to performance and energy efficiency enhancement of the inference phase of DNNs. The first one is model compression techniques while the second is efficient hardware implementation. Recent works on extremely-low-bit CNNs such as the binary neural network (BNN) and XNOR-Net replace the traditional floating-point operations with binary bit operations which significantly reduces the memory bandwidth and storage requirement. However, it suffers from non-negligible accuracy loss and underutilized digital signal processing (DSP) blocks of FPGAs. To overcome these limitations, this paper proposes REQ-YOLO, a resource-aware, systematic weight quantization framework for object detection, considering both algorithm and hardware resource aspects in object detection. We adopt the block-circulant matrix method and propose a heterogeneous weight quantization using the Alternating Direction Method of Multipliers (ADMM), an effective optimization technique for general, non-convex optimization problems. To achieve real-time, highly-efficient implementations on FPGA, we present the detailed hardware implementation of block circulant matrices on CONV layers and develop an efficient processing element (PE) structure supporting the heterogeneous weight quantization, CONV dataflow and pipelining techniques, design optimization, and a template-based automatic synthesis framework to optimally exploit hardware resource. Experimental results show that our proposed REQ-YOLO framework can significantly compress the YOLO model while introducing very small accuracy degradation.

cs.CV↗

arXiv2019

AutoCompress: An Automatic DNN Structured Pruning Framework for Ultra-High Compression Rates

Structured weight pruning is a representative model compression technique of DNNs to reduce the storage and computation requirements and accelerate inference. An automatic hyperparameter determination process is necessary due to the large number of flexible hyperparameters. This work proposes AutoCompress, an automatic structured pruning framework with the following key performance improvements: (i) effectively incorporate the combination of structured pruning schemes in the automatic process; (ii) adopt the state-of-art ADMM-based structured weight pruning as the core algorithm, and propose an innovative additional purification step for further weight reduction without accuracy loss; and (iii) develop effective heuristic search method enhanced by experience-based guided search, replacing the prior deep reinforcement learning technique which has underlying incompatibility with the target pruning problem. Extensive experiments on CIFAR-10 and ImageNet datasets demonstrate that AutoCompress is the key to achieve ultra-high pruning rates on the number of weights and FLOPs that cannot be achieved before. As an example, AutoCompress outperforms the prior work on automatic model compression by up to 33x in pruning rate (120x reduction in the actual parameter count) under the same accuracy. Significant inference speedup has been observed from the AutoCompress framework on actual measurements on smartphone. We release all models of this work at anonymous link: http://bit.ly/2VZ63dS.

cs.LG↗

arXiv2019

Distinguishing Individual Red Pandas from Their Faces

Individual identification is essential to animal behavior and ecology research and is of significant importance for protecting endangered species. Red pandas, among the world's rarest animals, are currently identified mainly by visual inspection and microelectronic chips, which are costly and inefficient. Motivated by recent advancement in computer-vision-based animal identification, in this paper, we propose an automatic framework for identifying individual red pandas based on their face images. We implement the framework by exploring well-established deep learning models with necessary adaptation for effectively dealing with red panda images. Based on a database of red panda images constructed by ourselves, we evaluate the effectiveness of the proposed automatic individual red panda identification method. The evaluation results show the promising potential of automatically recognizing individual red pandas from their faces. We are going to release our database and model in the public domain to promote the research on automatic animal identification and particularly on the technique for protecting red pandas.

cs.CV↗