arXiv subjects

Yue Liu

Publications and source records attributed to Yue Liu.

At least 325 records · Page 18Linked to original sources

arXiv2021

Personalized Entity Resolution with Dynamic Heterogeneous Knowledge Graph Representations

The growing popularity of Virtual Assistants poses new challenges for Entity Resolution, the task of linking mentions in text to their referent entities in a knowledge base. Specifically, in the shopping domain, customers tend to use implicit utterances (e.g., "organic milk") rather than explicit names, leading to a large number of candidate products. Meanwhile, for the same query, different customers may expect different results. For example, with "add milk to my cart", a customer may refer to a certain organic product, while some customers may want to re-order products they regularly purchase. To address these issues, we propose a new framework that leverages personalized features to improve the accuracy of product ranking. We first build a cross-source heterogeneous knowledge graph from customer purchase history and product knowledge graph to jointly learn customer and product embeddings. After that, we incorporate product, customer, and history representations into a neural reranking model to predict which candidate is most likely to be purchased for a specific customer. Experiments show that our model substantially improves the accuracy of the top ranked candidates by 24.6% compared to the state-of-the-art product search model.

cs.CL↗

arXiv2021

Energy efficiency of voltage waveform tailoring for the generation of excited species in RF plasma jets operated in He/N$_2$ mixtures

Based on Tunable Diode Laser Absorption Spectroscopy (TDLAS) measurements of the spatially averaged and peak helium metastable atom densities in a capacitively coupled micro atmospheric pressure plasma jet operated in He/N$_2$ mixtures, the energy efficiency of metastable species (He-I 2$^3$S$_1$) generation is compared for three different scenarios: single frequency operation at (i) 13.56 MHz and (ii) 54.12 MHz, and Voltage Waveform Tailoring (VWT) at (iii) "valleys"-waveforms synthesized from four consecutive harmonics of 13.56 MHz. For each case, the dissipated power is measured based on a careful calibration procedure of voltage and current measurements. The results are compared to PIC/MCC simulation results and very good agreement is found. The computational results show that the choice of the surface coefficients in the simulation is important to reproduce the experimental data correctly. Due to the enhanced control of the spatio-temporal electron power absorption dynamics and, thus, of the electron energy distribution function by VWT, this approach does not only provide better control of the generation of excited and reactive species compared to single frequency excitation, but in case of helium metastables the energy efficiency is also shown to be significantly higher in case of VWT.

physics.plasm-ph↗

arXiv2021

Option to survive or surrender: carbon asset management and optimization in thermal power enterprises from China

Carbon emission right allowance is a double-edged sword, one edge is to reduce emission as its original design intention, another edge has in practice slain many less developed coal-consuming enterprises, especially for those in thermal power industry. Partially governed on the hilt in hands of the authority, body of this sword is the prices of carbon emission right. How should the thermal power plants dance on the blade motivates this research. Considering the impact of price fluctuations of carbon emission right allowance, we investigate the operation of Chinese thermal power plant by modeling the decision-making with optimal stopping problem, which is established on the stochastic environment with carbon emission allowance price process simulated by geometric Brownian motion. Under the overall goal of maximizing the ultimate profitability, the optimal stopping indicates the timing of suspend or halt of production, hence the optimal stopping boundary curve implies the edge of life and death with regard to this enterprise. Applying this methodology, real cases of failure and survival of several Chinese representative thermal power plants were analyzed to explore the industry ecotope, which leads to the findings that: 1) The survival environment of existed thermal power plants becomes severer when facing more pressure from the newborn carbon-finance market. 2) Boundaries of survival environment is mainly drawn by the technical improvements for rising the utilization rate of carbon emission. Based on the same optimal stopping model, outlook of this industry is drawn with a demarcation surface defining the vivosphere of thermal power plants with different levels of profitability. This finding provides benchmarks for those enterprises struggling for survival and policy makers scheming better supervision and necessary intervene.

econ.GN↗

arXiv2021

Risk Variance Penalization

The key of the out-of-distribution (OOD) generalization is to generalize invariance from training domains to target domains. The variance risk extrapolation (V-REx) is a practical OOD method, which depends on a domain-level regularization but lacks theoretical verifications about its motivation and utility. This article provides theoretical insights into V-REx by studying a variance-based regularizer. We propose Risk Variance Penalization (RVP), which slightly changes the regularization of V-REx but addresses the theory concerns about V-REx. We provide theoretical explanations and a theory-inspired tuning scheme for the regularization parameter of RVP. Our results point out that RVP discovers a robust predictor. Finally, we experimentally show that the proposed regularizer can find an invariant predictor under certain conditions.

cs.LG↗

arXiv2021

Beyond Visual Attractiveness: Physically Plausible Single Image HDR Reconstruction for Spherical Panoramas

HDR reconstruction is an important task in computer vision with many industrial needs. The traditional approaches merge multiple exposure shots to generate HDRs that correspond to the physical quantity of illuminance of the scene. However, the tedious capturing process makes such multi-shot approaches inconvenient in practice. In contrast, recent single-shot methods predict a visually appealing HDR from a single LDR image through deep learning. But it is not clear whether the previously mentioned physical properties would still hold, without training the network to explicitly model them. In this paper, we introduce the physical illuminance constraints to our single-shot HDR reconstruction framework, with a focus on spherical panoramas. By the proposed physical regularization, our method can generate HDRs which are not only visually appealing but also physically plausible. For evaluation, we collect a large dataset of LDR and HDR images with ground truth illuminance measures. Extensive experiments show that our HDR images not only maintain high visual quality but also top all baseline methods in illuminance prediction accuracy.

cs.CV↗

arXiv2021

Electron heating mode transitions in radio-frequency driven micro atmospheric pressure plasma jets in He/O$_{2}$: A fluid dynamics approach

A two-dimensional fluid model is used to investigate the electron heating dynamics and the production of neutral species in a capacitively coupled radio-frequency micro atmospheric pressure helium plasma jet -- specifically the COST jet -- with a small oxygen admixture. Electron heating mode transitions are found to be induced by varying the driving voltage amplitude and the O$_2$ concentration numerically and experimentally. The helium metastable density, and the charged species densities are highly relevant to the electron heating dynamics. By analyzing the creation and destruction mechanisms of the negative ions, we find that the generation of negative ions strongly depends on the O$_2$ concentration. The increase of the electronegativity with the increasing O$_2$ concentration leads to an enhancement of the bulk drift electric field. The distributions of the different neutral species densities along the direction of the gas flow inside the jet, as well as in the effluent differ a lot due to the relevant chemical reaction rates and the effect of the gas flow. The simulated results show that a fluid model can be an effective tool for qualitative investigations of micro atmospheric pressure plasma jets.

physics.plasm-ph↗

arXiv2021

Out-of-Distribution Generalization Analysis via Influence Function

The mismatch between training and target data is one major challenge for current machine learning systems. When training data is collected from multiple domains and the target domains include all training domains and other new domains, we are facing an Out-of-Distribution (OOD) generalization problem that aims to find a model with the best OOD accuracy. One of the definitions of OOD accuracy is worst-domain accuracy. In general, the set of target domains is unknown, and the worst over target domains may be unseen when the number of observed domains is limited. In this paper, we show that the worst accuracy over the observed domains may dramatically fail to identify the OOD accuracy. To this end, we introduce Influence Function, a classical tool from robust statistics, into the OOD generalization problem and suggest the variance of influence function to monitor the stability of a model on training domains. We show that the accuracy on test domains and the proposed index together can help us discern whether OOD algorithms are needed and whether a model achieves good OOD generalization.

cs.LG↗

arXiv2020

Electron pairing in the pseudogap state revealed by shot noise in copper-oxide junctions

In the quest to understand high-temperature superconductivity in copper oxides, a vigorous debate has been focused on the pseudogap - a partial gap that opens over portions of the Fermi surface in the 'normal' state above the bulk critical temperature ($T_{c}$). The pseudogap has been attributed to precursor superconductivity, to the existence of preformed pairs, or to competing orders such as charge-density waves. A direct determination of the charge of carriers as a function of temperature and bias could help resolve among these alternatives. Here, we report measurements of the shot noise of tunneling current in high-quality La$_{2-x}$Sr$_{x}$CuO$_{4}$/La$_{2}$CuO$_{4}$/La$_{2-x}$Sr$_{x}$CuO$_{4}$ (LSCO/LCO/LSCO) heterostructures fabricated using atomic-layer-by-layer molecular beam epitaxy, for several doping levels. The data delineate three distinct regions in the bias voltage-temperature ($V-T$) space. Well outside the superconducting gap region, the shot noise agrees quantitatively with independent tunneling of charge-e carriers. Deep within the gap, shot noise is greatly enhanced, reminiscent of multiple Andreev reflections. Starting above $T_{c}$ and extending to biases much larger than the gap, there is a broad region in which the noise substantially exceeds the expectations of single-charge tunneling, indicating pairing of carriers. Pairs are detectable deep into the pseudogap region of temperature and bias.

cond-mat.supr-con↗

arXiv2020

Reconstruction Condition of Quantized Signals in Unlimited Sampling Framework

The latest theoretical advances in the field of unlimited sampling framework (USF) show the potential to avoid clipping problems of analog-to-digital converters (ADC). To date, most of the related works have focused on real-valued modulo samples, but little has been reported about the impact of quantization. In this paper, we study more practical USF system where modulo samples are quantized to a finite number of bits. In particular, we present a new requirement about the lower bound of sampling rate to ensure exact recovery of original signals from quantized modulo samples. The minimum sampling rate is jointly determined by signal bandwidth and quantization bits. Numerical results show that in the presence of quantization noise, signals with different waveforms and bandwidths are recovered perfectly at the new minimum sampling rate while the recovery fails at minimum sampling rate before modification, which also verifies the correctness of the theory. The trade-offs of sampling rates, quantization bits and computational complexity of recovery algorithm are also given for practitioners to weigh.

eess.SP↗

arXiv2020

Blockchain-based Federated Learning for Device Failure Detection in Industrial IoT

Device failure detection is one of most essential problems in industrial internet of things (IIoT). However, in conventional IIoT device failure detection, client devices need to upload raw data to the central server for model training, which might lead to disclosure of sensitive business data. Therefore, in this paper, to ensure client data privacy, we propose a blockchain-based federated learning approach for device failure detection in IIoT. First, we present a platform architecture of blockchain-based federated learning systems for failure detection in IIoT, which enables verifiable integrity of client data. In the architecture, each client periodically creates a Merkle tree in which each leaf node represents a client data record, and stores the tree root on a blockchain. Further, to address the data heterogeneity issue in IIoT failure detection, we propose a novel centroid distance weighted federated averaging (CDW\_FedAvg) algorithm taking into account the distance between positive class and negative class of each client dataset. In addition, to motivate clients to participate in federated learning, a smart contact based incentive mechanism is designed depending on the size and the centroid distance of client data used in local model training. A prototype of the proposed architecture is implemented with our industry partner, and evaluated in terms of feasibility, accuracy and performance. The results show that the approach is feasible, and has satisfactory accuracy and performance.

cs.DC↗

arXiv2020

Decoupling Inherent Risk and Early Cancer Signs in Image-based Breast Cancer Risk Models

The ability to accurately estimate risk of developing breast cancer would be invaluable for clinical decision-making. One promising new approach is to integrate image-based risk models based on deep neural networks. However, one must take care when using such models, as selection of training data influences the patterns the network will learn to identify. With this in mind, we trained networks using three different criteria to select the positive training data (i.e. images from patients that will develop cancer): an inherent risk model trained on images with no visible signs of cancer, a cancer signs model trained on images containing cancer or early signs of cancer, and a conflated model trained on all images from patients with a cancer diagnosis. We find that these three models learn distinctive features that focus on different patterns, which translates to contrasts in performance. Short-term risk is best estimated by the cancer signs model, whilst long-term risk is best estimated by the inherent risk model. Carelessly training with all images conflates inherent risk with early cancer signs, and yields sub-optimal estimates in both regimes. As a consequence, conflated models may lead physicians to recommend preventative action when early cancer signs are already visible.

eess.IV↗

arXiv2020

Deep Neural Network Approach for Annual Luminance Simulations

Annual luminance maps provide meaningful evaluations for occupants' visual comfort, preferences, and perception. However, acquiring long-term luminance maps require labor-intensive and time-consuming simulations or impracticable long-term field measurements. This paper presents a novel data-driven machine learning approach that makes annual luminance-based evaluations more efficient and accessible. The methodology is based on predicting the annual luminance maps from a limited number of point-in-time high dynamic range imagery by utilizing a deep neural network (DNN). Panoramic views are utilized, as they can be post-processed to study multiple view directions. The proposed DNN model can faithfully predict high-quality annual panoramic luminance maps from one of the three options within 30 minutes training time: a) point-in-time luminance imagery spanning 5% of the year, when evenly distributed during daylight hours, b) one-month hourly imagery generated or collected continuously during daylight hours around the equinoxes (8% of the year); or c) 9 days of hourly data collected around the spring equinox, summer and winter solstices (2.5% of the year) all suffice to predict the luminance maps for the rest of the year. The DNN predicted high-quality panoramas are validated against Radiance (RPICT) renderings using a series of quantitative and qualitative metrics. The most efficient predictions are achieved with 9 days of hourly data collected around the spring equinox, summer and winter solstices. The results clearly show that practitioners and researchers can efficiently incorporate long-term luminance-based metrics over multiple view directions into the design and research processes using the proposed DNN workflow.

cs.CV↗

arXiv2020

Video Moment Retrieval via Natural Language Queries

In this paper, we propose a novel method for video moment retrieval (VMR) that achieves state of the arts (SOTA) performance on R@1 metrics and surpassing the SOTA on the high IoU metric (R@1, IoU=0.7). First, we propose to use a multi-head self-attention mechanism, and further a cross-attention scheme to capture video/query interaction and long-range query dependencies from video context. The attention-based methods can develop frame-to-query interaction and query-to-frame interaction at arbitrary positions and the multi-head setting ensures the sufficient understanding of complicated dependencies. Our model has a simple architecture, which enables faster training and inference while maintaining . Second, We also propose to use multiple task training objective consists of moment segmentation task, start/end distribution prediction and start/end location regression task. We have verified that start/end prediction are noisy due to annotator disagreement and joint training with moment segmentation task can provide richer information since frames inside the target clip are also utilized as positive training examples. Third, we propose to use an early fusion approach, which achieves better performance at the cost of inference time. However, the inference time will not be a problem for our model since our model has a simple architecture which enables efficient training and inference.

cs.CV↗

arXiv2020

A Blockchain-based Platform Architecture for Multimedia Data Management

Massive amounts of multimedia data (i.e., text, audio, video, graphics and animation) are being generated everyday. Conventionally, multimedia data are managed by the platforms maintained by multimedia service providers, which are generally designed using centralised architecture. However, such centralised architecture may lead to a single point of failure and disputes over royalties or other rights. It is hard to ensure the data integrity and track fulfilment of obligations listed on the copyright agreement. To tackle these issues, in this paper, we present a blockchain-based platform architecture for multimedia data management. We adopt self-sovereign identity for identity management and design a multi-level capability-based mechanism for access control. We implement a proof-of-concept prototype using the proposed approach and evaluate it using a use case. The results show that the proposed approach is feasible and has scalable performance.

cs.CR↗

arXiv2020

Adding Seemingly Uninformative Labels Helps in Low Data Regimes

Evidence suggests that networks trained on large datasets generalize well not solely because of the numerous training examples, but also class diversity which encourages learning of enriched features. This raises the question of whether this remains true when data is scarce - is there an advantage to learning with additional labels in low-data regimes? In this work, we consider a task that requires difficult-to-obtain expert annotations: tumor segmentation in mammography images. We show that, in low-data settings, performance can be improved by complementing the expert annotations with seemingly uninformative labels from non-expert annotators, turning the task into a multi-class problem. We reveal that these gains increase when less expert data is available, and uncover several interesting properties through further studies. We demonstrate our findings on CSAW-S, a new dataset that we introduce here, and confirm them on two public datasets.

cs.CV↗

arXiv2020

Online NEAT for Credit Evaluation -- a Dynamic Problem with Sequential Data

In this paper, we describe application of Neuroevolution to a P2P lending problem in which a credit evaluation model is updated based on streaming data. We apply the algorithm Neuroevolution of Augmenting Topologies (NEAT) which has not been widely applied generally in the credit evaluation domain. In addition to comparing the methodology with other widely applied machine learning techniques, we develop and evaluate several enhancements to the algorithm which make it suitable for the particular aspects of online learning that are relevant in the problem. These include handling unbalanced streaming data, high computation costs, and maintaining model similarity over time, that is training the stochastic learning algorithm with new data but minimizing model change except where there is a clear benefit for model performance

cs.LG↗

arXiv2020

Orbital Stability of smooth solitary waves for the Degasperis-Procesi Equation

The Degasperis-Procesi equation is the integrable Camassa-Holm-type model which is an asymptotic approximation for the unidirectional propagation of shallow water waves. This work establishes the orbital stability of localized smooth solitary waves to the Desgasperis-Procesi (DP) equation on the real line. %extending our previous work on their spectral stability \cite{LLW}. The main difficulty stems from the fact that the translation symmetry for the DP equation gives rise to a conserved quantity equivalent to the $L^2$-norm, which by itself can not bound the higher-order nonlinear terms in the Lagrangian. The remedy is to observe that, given a sufficiently smooth initial condition satisfying a measurable constraint, the $L^\infty$ orbital norm of the perturbation is bounded above by a function of its $L^2$ orbital norm, yielding the orbital stability in the $L^2\cap L^\infty$ space.

math.AP↗

arXiv2020

Design Patterns for Blockchain-based Self-Sovereign Identity

Self-sovereign identity is a new identity management paradigm that allows entities to really have the ownership of their identity data and control their use without involving any intermediary. Blockchain is an enabling technology for building self-sovereign identity systems by providing a neutral and trustable storage and computing infrastructure and can be viewed as a component of the systems. Both blockchain and self-sovereign identity are emerging technologies which could present a steep learning curve for architects. We collect and propose 12 design patterns for blockchain-based self-sovereign identity systems to help the architects understand and easily apply the concepts in system design. Based on the lifecycles of three main objects involved in self-sovereign identity, we categorise the patterns into three groups: key management patterns, decentralised identifier management patterns, and credential design patterns. The proposed patterns provide a systematic and holistic guide for architects to design the architecture of blockchain-based self-sovereign identity systems.

cs.CR↗