arXiv subjects

Ning Ding

Publications and source records attributed to Ning Ding.

At least 199 records · Page 11Linked to original sources

arXiv2021

Maximize the Foot Clearance for a Hopping Robotic Leg Considering Motor Saturation

A hopping leg, no matter in legged animals or humans, usually behaves like a spring during the periodic hopping. Hopping like a spring is efficient and without the requirement of complicated control algorithms. Position and force control are two main methods to realize such a spring-like behaviour. The position control usually consumes the torque resources to ensure the position accuracy and compensate the tracking errors. In comparison, the force control strategy is able to maintain a high elasticity. Currently, the position and force control both leads to the discount of motor saturation ratio as well as the bandwidth of the control system, and thus attenuates the performance of the actuator. To augment the performance, this letter proposes a motor saturation strategy based on the force control to maximize the output torque of the actuator and realize the continuous hopping motion with natural dynamics. The proposed strategy is able to maximize the saturation ratio of motor and thus maximize the foot clearance of the single leg. The dynamics of the two-mass model is utilized to increase the force bandwidth and the performance of the actuator. A single leg with two degrees of freedom is designed as the experiment platform. The actuator consists of a powerful electric motor, a harmonic gear and encoder. The effectiveness of this method is verified through simulations and experiments using a robotic leg actuated by powerful high reduction ratio actuators.

cs.RO↗

arXiv2021

Discriminative-Generative Representation Learning for One-Class Anomaly Detection

As a kind of generative self-supervised learning methods, generative adversarial nets have been widely studied in the field of anomaly detection. However, the representation learning ability of the generator is limited since it pays too much attention to pixel-level details, and generator is difficult to learn abstract semantic representations from label prediction pretext tasks as effective as discriminator. In order to improve the representation learning ability of generator, we propose a self-supervised learning framework combining generative methods and discriminative methods. The generator no longer learns representation by reconstruction error, but the guidance of discriminator, and could benefit from pretext tasks designed for discriminative methods. Our discriminative-generative representation learning method has performance close to discriminative methods and has a great advantage in speed. Our method used in one-class anomaly detection task significantly outperforms several state-of-the-arts on multiple benchmark data sets, increases the performance of the top-performing GAN-based baseline by 6% on CIFAR-10 and 2% on MVTAD.

cs.CV↗

arXiv2021

CLINE: Contrastive Learning with Semantic Negative Examples for Natural Language Understanding

Despite pre-trained language models have proven useful for learning high-quality semantic representations, these models are still vulnerable to simple perturbations. Recent works aimed to improve the robustness of pre-trained models mainly focus on adversarial training from perturbed examples with similar semantics, neglecting the utilization of different or even opposite semantics. Different from the image processing field, the text is discrete and few word substitutions can cause significant semantic changes. To study the impact of semantics caused by small perturbations, we conduct a series of pilot experiments and surprisingly find that adversarial training is useless or even harmful for the model to detect these semantic changes. To address this problem, we propose Contrastive Learning with semantIc Negative Examples (CLINE), which constructs semantic negative examples unsupervised to improve the robustness under semantically adversarial attacking. By comparing with similar and opposite semantic examples, the model can effectively perceive the semantic changes caused by small perturbations. Empirical results show that our approach yields substantial improvements on a range of sentiment analysis, reasoning, and reading comprehension tasks. And CLINE also ensures the compactness within the same semantics and separability across different semantics in sentence-level.

cs.CL↗

arXiv2021

Peierls transition driven ferroelasticity in two-dimensional $d$-$f$ hybrid magnet

For broad nanoscale applications, it is crucial to implement more functional properties, especially those ferroic orders, into two-dimensional materials. Here GdI$_3$ is theoretically identified as a honeycomb antiferromagnet with large $4f$ magnetic moment. The intercalation of metal atoms can dope electrons into Gd's $5d$-orbitals, which alters its magnetic state and lead to Peierls transition. Due to the strong electron-phonon coupling, the Peierls transition induces prominent ferroelasticity, making it a multiferroic system. The strain from undirectional stretching can be self-relaxed via resizing of triple ferroelastic domains, which can protect the magnet aganist mechnical breaking in flexible applications.

cond-mat.str-el↗

arXiv2021

Prototypical Representation Learning for Relation Extraction

Recognizing relations between entities is a pivotal task of relational learning. Learning relation representations from distantly-labeled datasets is difficult because of the abundant label noise and complicated expressions in human language. This paper aims to learn predictive, interpretable, and robust relation representations from distantly-labeled data that are effective in different settings, including supervised, distantly supervised, and few-shot learning. Instead of solely relying on the supervision from noisy labels, we propose to learn prototypes for each relation from contextual information to best explore the intrinsic semantics of relations. Prototypes are representations in the feature space abstracting the essential semantics of relations between entities in sentences. We learn prototypes based on objectives with clear geometric interpretation, where the prototypes are unit vectors uniformly dispersed in a unit ball, and statement embeddings are centered at the end of their corresponding prototype vectors on the surface of the ball. This approach allows us to learn meaningful, interpretable prototypes for the final classification. Results on several relation learning tasks show that our model significantly outperforms the previous state-of-the-art models. We further demonstrate the robustness of the encoder and the interpretability of prototypes with extensive experiments.

cs.CL↗

arXiv2021

A Hybrid Task-Oriented Dialog System with Domain and Task Adaptive Pretraining

This paper describes our submission for the End-to-end Multi-domain Task Completion Dialog shared task at the 9th Dialog System Technology Challenge (DSTC-9). Participants in the shared task build an end-to-end task completion dialog system which is evaluated by human evaluation and a user simulator based automatic evaluation. Different from traditional pipelined approaches where modules are optimized individually and suffer from cascading failure, we propose an end-to-end dialog system that 1) uses Generative Pretraining 2 (GPT-2) as the backbone to jointly solve Natural Language Understanding, Dialog State Tracking, and Natural Language Generation tasks, 2) adopts Domain and Task Adaptive Pretraining to tailor GPT-2 to the dialog domain before finetuning, 3) utilizes heuristic pre/post-processing rules that greatly simplify the prediction tasks and improve generalizability, and 4) equips a fault tolerance module to correct errors and inappropriate responses. Our proposed method significantly outperforms baselines and ties for first place in the official evaluation. We make our source code publicly available.

cs.CL↗

arXiv2020

Ferroelectricity and ferromagnetism in VOI$_2$ monolayer: the role of Dzyaloshinskii-Moriya interaction

Multiferroics with intrinsic ferromagnetism and ferroelectricity are highly desired but rather rare, while most ferroelectric magnets are antiferromagnetic. A recent theoretical work [Phys. Rev. B {\bf 99}, 195434 (2019)] predicted that oxyhalides VO$X_2$ ($X$: halogen) monolayers are two-dimensional multiferroics by violating the empirical $d^0$ rule. Most interestingly, the member VOI$_2$ are predicted to exhibit spontaneous ferromagnetism and ferroelectricity. In this work, we extend the previous study on the structure and magnetism of VOI$_2$ monolayer by using density functional theory and Monte Carlo simulation. The presence of the heavy element iodine with a strong spin-orbit coupling leads an effective Dzyaloshinskii-Moriya interaction in the polar structure, which favors a short-period spiral a magnetic structure.. Another interesting result is that the on-site Coulomb interaction can strongly suppress the polar distortion thus leading to a ferromagnetic metallic state. Therefore, the VOI2 monolayer is either a ferroelectric insulator with spiral magnetism or a ferromagnetic metal, instead of a ferromagnetic ferroelectric system. Our study highlights the key physical role of the Dzyaloshinskii-Moriya interaction.

cond-mat.mtrl-sci↗

arXiv2020

Prediction of two-dimensional ferromagnetic ferroelectric VOF$_2$ monolayer

Nowadays, designing and searching for materials with multiple functional characteristics are the key to achieving high-performance electronic devices. Among many candidates, two-dimensional multiferroic materials have great potential to be applied in highly-integrated magnetoelectric devices, such as high-density non-volatile memories. Here, we predict a two-dimensional material VOF2 monolayer to possess intrinsic ferroelectric and ferromagnetic properties. The VOF$_2$ monolayer own the largest in-plane ferroelectric polarization ($332$ pC/m) in the family of VO$X_2$ ($X$: halogen) oxyhalides. Different from other VO$X_2$ monolayers whose magnetic ground states are antiferromagnetic or noncollinear spiral textures, the VOF$_2$ monolayer owns a robust ferromagnetic ground state, which is rare but highly desirable. Our theoretical prediction provides a good candidate and starting point for the further pursuit of more two-dimensional multiferroic materials with high-performance magnetoelectricity.

cond-mat.mtrl-sci↗

arXiv2020

TP-LSD: Tri-Points Based Line Segment Detector

This paper proposes a novel deep convolutional model, Tri-Points Based Line Segment Detector (TP-LSD), to detect line segments in an image at real-time speed. The previous related methods typically use the two-step strategy, relying on either heuristic post-process or extra classifier. To realize one-step detection with a faster and more compact model, we introduce the tri-points representation, converting the line segment detection to the end-to-end prediction of a root-point and two endpoints for each line segment. TP-LSD has two branches: tri-points extraction branch and line segmentation branch. The former predicts the heat map of root-points and the two displacement maps of endpoints. The latter segments the pixels on straight lines out from background. Moreover, the line segmentation map is reused in the first branch as structural prior. We propose an additional novel evaluation metric and evaluate our method on Wireframe and YorkUrban datasets, demonstrating not only the competitive accuracy compared to the most recent methods, but also the real-time run speed up to 78 FPS with the $320\times 320$ input.

cs.CV↗

arXiv2020

Coupling Distant Annotation and Adversarial Training for Cross-Domain Chinese Word Segmentation

Fully supervised neural approaches have achieved significant progress in the task of Chinese word segmentation (CWS). Nevertheless, the performance of supervised models tends to drop dramatically when they are applied to out-of-domain data. Performance degradation is caused by the distribution gap across domains and the out of vocabulary (OOV) problem. In order to simultaneously alleviate these two issues, this paper proposes to couple distant annotation and adversarial training for cross-domain CWS. For distant annotation, we rethink the essence of "Chinese words" and design an automatic distant annotation mechanism that does not need any supervision or pre-defined dictionaries from the target domain. The approach could effectively explore domain-specific words and distantly annotate the raw texts for the target domain. For adversarial training, we develop a sentence-level training procedure to perform noise reduction and maximum utilization of the source domain information. Experiments on multiple real-world datasets across various domains show the superiority and robustness of our model, significantly outperforming previous state-of-the-art cross-domain CWS methods.

cs.CL↗

arXiv2020

Length-Controllable Image Captioning

The last decade has witnessed remarkable progress in the image captioning task; however, most existing methods cannot control their captions, \emph{e.g.}, choosing to describe the image either roughly or in detail. In this paper, we propose to use a simple length level embedding to endow them with this ability. Moreover, due to their autoregressive nature, the computational complexity of existing models increases linearly as the length of the generated captions grows. Thus, we further devise a non-autoregressive image captioning approach that can generate captions in a length-irrelevant complexity. We verify the merit of the proposed length level embedding on three models: two state-of-the-art (SOTA) autoregressive models with different types of decoder, as well as our proposed non-autoregressive model, to show its generalization ability. In the experiments, our length-controllable image captioning models not only achieve SOTA performance on the challenging MS COCO dataset but also generate length-controllable and diverse image captions. Specifically, our non-autoregressive model outperforms the autoregressive baselines in terms of controllability and diversity, and also significantly improves the decoding efficiency for long captions. Our code and models are released at \textcolor{magenta}{\texttt{https://github.com/bearcatt/LaBERT}}.

cs.CV↗

arXiv2019

DELTA: A DEep learning based Language Technology plAtform

In this paper we present DELTA, a deep learning based language technology platform. DELTA is an end-to-end platform designed to solve industry level natural language and speech processing problems. It integrates most popular neural network models for training as well as comprehensive deployment tools for production. DELTA aims to provide easy and fast experiences for using, deploying, and developing natural language processing and speech models for both academia and industry use cases. We demonstrate the reliable performance with DELTA on several natural language processing and speech tasks, including text classification, named entity recognition, natural language inference, speech recognition, speaker verification, etc. DELTA has been used for developing several state-of-the-art algorithms for publications and delivering real production to serve millions of users.

cs.CL↗

arXiv2019

Wireless Key Generation from Imperfect Channel State Information: Performance Analysis and Improvements

The basis of generating secret key from the common wireless channel at two communication parties is reciprocity. However, due to non-simultaneous measurements and unavoidable estimation error, bias always exists and key disagreement occurs. In this paper, performance analysis and improvements for key generation exploiting imperfect channel statement information (CSI) is studied. Firstly, the closed-form expression for key disagreement rate (KDR) is derived to evaluate the key mismatch caused by time delay and estimation error. Secondly, the energy efficiency involving power allocation between key generation and private data transmission is presented. After that, a neural network based prediction (NNBP) algorithm is proposed to improve the key consistency. Simulation results verify the theoretical analysis and show that NNBP achieves significant improvements. And it's interesting to find that NNBP algorithm achieves a higher performance gain in low signal-to-noise ration (SNR) and high delay condition, which is different from other methods, like guard band based quantizer (GBBQ).

cs.IT↗

arXiv2005

The Economic Mobility in Money Transfer

In this paper, we investigate the economic mobility in some money transfer models which have been applied into the research on wealth distribution. We demonstrate the mobility by recording the time series of agents' ranks and observing their volatility. We also compare the mobility quantitatively by employing an index, "the per capita aggregate change in log-income", raised by economists. Like the shape of distribution, the character of mobility is also decided by the trading rule in these transfer models. It is worth noting that even though different models have the same type of distribution, their mobility characters may be quite different.

physics.soc-ph↗

arXiv2005

Prospects for Money Transfer Models

Recently, in order to explore the mechanism behind wealth or income distribution, several models have been proposed by applying principles of statistical mechanics. These models share some characteristics, such as consisting of a group of individual agents, a pile of money and a specific trading rule. Whatever the trading rule is, the most noteworthy fact is that money is always transferred from one agent to another in the transferring process. So we call them money transfer models. Besides explaining income and wealth distributions, money transfer models can also be applied to other disciplines. In this paper we summarize these areas as statistical distribution, economic mobility, transfer rate and money creation. First, money distribution (or income distribution) can be exhibited by recording the money stock (flow). Second, the economic mobility can be shown by tracing the change in wealth or income over time for each agent. Third, the transfer rate of money and its determinants can be analyzed by tracing the transferring process of each one unit of money. Finally, money creation process can also be investigated by permitting agents go into debts. Some future extensions to these models are anticipated to be structural improvement and generalized mathematical analysis.

physics.soc-ph↗

arXiv2005

Dynamic Process of Money Transfer Models

We have studied numerically the statistical mechanics of the dynamic phenomena, including money circulation and economic mobility, in some transfer models. The models on which our investigations were performed are the basic model proposed by A. Dragulescu and V. Yakovenko [1], the model with uniform saving rate developed by A. Chakraborti and B.K. Chakrabarti [2], and its extended model with diverse saving rate [3]. The velocity of circulation is found to be inversely related with the average holding time of money. In order to check the nature of money transferring process in these models, we demonstrated the probability distributions of holding time. In the model with uniform saving rate, the distribution obeys exponential law, which indicates money transfer here is a kind of Poisson process. But when the saving rate is set diversely, the holding time distribution follows a power law. The velocity can also be deduced from a typical individual's optimal choice. In this way, an approach for building the micro-foundation of velocity is provided. In order to expose the dynamic mechanism behind the distribution in microscope, we examined the mobility by collecting the time series of agents' rank and measured it by employing an index raised by economists. In the model with uniform saving rate, the higher saving rate, the slower agents moves in the economy. Meanwhile, all of the agents have the same chance to be the rich. However, it is not the case in the model with diverse saving rate, where the assumed economy falls into stratification. The volatility distribution of the agents' ranks are also demonstrated to distinguish the differences among these models.

physics.soc-ph↗

arXiv2005

How Required Reserve Ratio Affects Distribution and Velocity of Money

In this paper the dependence of wealth distribution and the velocity of money on the required reserve ratio is examined based on a random transfer model of money and computer simulations. A fractional reserve banking system is introduced to the model where money creation can be achieved by bank loans and the monetary aggregate is determined by the monetary base and the required reserve ratio. It is shown that monetary wealth follows asymmetric Laplace distribution and latency time of money follows exponential distribution. The expression of monetary wealth distribution and that of the velocity of money in terms of the required reserve ratio are presented in a good agreement with simulation results.

physics.soc-ph↗

arXiv2005

The Circulation of Money and Holding Time Distribution

We have studied the statistical mechanics of money circulation in a closed economic system. An explicit statistical formulation of the circulation velocity of money is presented for the first time by introducing the concept of holding time of money. The result indicates that the velocity is governed by behavior patterns of economic agents. Computer simulations have been carried out in order to demonstrate the shape of the holding time distribution. We find that, money circulation is a Poisson process in which the holding time probability distribution follows a type of Gamma distribution, and the velocity of money depends on the share for exchange and the number of agents.

physics.soc-ph↗