arXiv subjects

Di Wu

Publications and source records attributed to Di Wu.

At least 505 records · Page 28Linked to original sources

arXiv2023

Meta-Learning Based Early Fault Detection for Rolling Bearings via Few-Shot Anomaly Detection

Early fault detection (EFD) of rolling bearings can recognize slight deviation of the health states and contribute to the stability of mechanical systems. In practice, very limited target bearing data are available to conduct EFD, which makes it hard to adapt to the EFD task of new bearings. To address this problem, many transfer learning based EFD methods utilize historical data to learn transferable domain knowledge and conduct early fault detection on new target bearings. However, most existing methods only consider the distribution drift across different working conditions but ignore the difference between bearings under the same working condition, which is called Unit-to-Unit Variability (UtUV). The setting of EFD with limited target data considering UtUV can be formulated as a Few-shot Anomaly Detection task. Therefore, this paper proposes a novel EFD method based on meta-learning considering UtUV. The proposed method can learn a generic metric based on Relation Network (RN) to measure the similarity between normal data and the new arrival target bearing data. Besides, the proposed method utilizes a health state embedding strategy to decrease false alarms. The performance of proposed method is tested on two bearing datasets. The results show that the proposed method can detect incipient faults earlier than the baselines with lower false alarms.

cs.LG↗

arXiv2023

An Early Fault Detection Method of Rotating Machines Based on Multiple Feature Fusion with Stacking Architecture

Early fault detection (EFD) of rotating machines is important to decrease the maintenance cost and improve the mechanical system stability. One of the key points of EFD is developing a generic model to extract robust and discriminative features from different equipment for early fault detection. Most existing EFD methods focus on learning fault representation by one type of feature. However, a combination of multiple features can capture a more comprehensive representation of system state. In this paper, we propose an EFD method based on multiple feature fusion with stacking architecture (M2FSA). The proposed method can extract generic and discriminiative features to detect early faults by combining time domain (TD), frequency domain (FD), and time-frequency domain (TFD) features. In order to unify the dimensions of the different domain features, Stacked Denoising Autoencoder (SDAE) is utilized to learn deep features in three domains. The architecture of the proposed M2FSA consists of two layers. The first layer contains three base models, whose corresponding inputs are different deep features. The outputs of the first layer are concatenated to generate the input to the second layer, which consists of a meta model. The proposed method is tested on three bearing datasets. The results demonstrate that the proposed method is better than existing methods both in sensibility and reliability.

cs.LG↗

arXiv2023

Negative electrocaloric effect in nonpolar phases of perovskite over wide range of temperature

The electrocaloric effect (ECE) offers a promising alternative to the traditional gas compressing refrigeration due to its high efficiency and environmental friendliness. The unusual negative electrocaloric effect refers to the adiabatic temperature drops due to application of electric field, in contrast with the normal (positive) ECE, and provides ways to improve the electrocaloric efficiency in refrigeration cycles. However, negative ECE is unusual and requires a clear understanding of microscopic mechanisms. Here, we found unexpected and extensive negative ECE in nonpolar orthorhombic, tetragonal, and cubic phases of halide and oxide perovskite at wide range of temperature by means of first-principle-based large scale Monte Carlo methods. Such unexpected negative ECE originates from the octahedral tilting related entropy change rather than the polarization entropy change under the application of electric field. Furthermore, a giant negative ECE with temperature change of 8.6 K is found at room temperature. This giant and extensive negative ECE in perovskite opens up new horizon in the research of caloric effects and broadens the electrocaloric refrigeration ways with high efficiency.

cond-mat.mtrl-sci↗

arXiv2023

A Modified Sequence-to-point HVAC Load Disaggregation Algorithm

This paper presents a modified sequence-to-point (S2P) algorithm for disaggregating the heat, ventilation, and air conditioning (HVAC) load from the total building electricity consumption. The original S2P model is convolutional neural network (CNN) based, which uses load profiles as inputs. We propose three modifications. First, the input convolution layer is changed from 1D to 2D so that normalized temperature profiles are also used as inputs to the S2P model. Second, a drop-out layer is added to improve adaptability and generalizability so that the model trained in one area can be transferred to other geographical areas without labelled HVAC data. Third, a fine-tuning process is proposed for areas with a small amount of labelled HVAC data so that the pre-trained S2P model can be fine-tuned to achieve higher disaggregation accuracy (i.e., better transferability) in other areas. The model is first trained and tested using smart meter and sub-metered HVAC data collected in Austin, Texas. Then, the trained model is tested on two other areas: Boulder, Colorado and San Diego, California. Simulation results show that the proposed modified S2P algorithm outperforms the original S2P model and the support-vector machine based approach in accuracy, adaptability, and transferability.

eess.SY↗

arXiv2023

Online Sparse Streaming Feature Selection Using Adapted Classification

Traditional feature selections need to know the feature space before learning, and online streaming feature selection (OSFS) is proposed to process streaming features on the fly. Existing methods divide features into relevance or irrelevance without missing data, and deleting irrelevant features may lead to in-formation loss. Motivated by this, we focus on completing the streaming feature matrix and division of feature correlation and propose online sparse streaming feature selection based on adapted classification (OS2FS-AC). This study uses Latent Factor Analysis (LFA) to pre-estimate missed data. Besides, we use the adaptive method to obtain the threshold, divide the features into strongly relevant, weakly relevant, and irrelevant features, and then divide weak relevance with more information. Experimental results on ten real-world data sets demonstrate that OS2FS-AC performs better than state-of-the-art algo-rithms.

cs.LG↗

arXiv2023

Structure and coloring of some ($P_7,C_4$)-free graphs

Let $G$ be a graph. We use $P_t$ and $C_t$ to denote a path and a cycle on $t$ vertices, respectively. A {\em diamond} is a graph obtained from two triangles that share exactly one edge. A {\em kite} is a graph consists of a diamond and another vertex adjacent to a vertex of degree 2 of the diamond. A {\em gem} is a graph that consists of a $P_4$ plus a vertex adjacent to all vertices of the $P_4$. In this paper, we prove some structural properties to $(P_7, C_4,$ diamond)-free graphs, $(P_7, C_4,$ kite)-free graphs and $(P_7, C_4,$ gem)-free graphs. As their corollaries, we show that (\romannumeral 1) $χ(G)\leq \max\{3,ω(G)\}$ if $G$ is $(P_7, C_4,$ diamond)-free, (\romannumeral 2) $χ(G)\leq ω(G)+1$ if $G$ is $(P_7, C_4,$ kite)-free and (\romannumeral 3) $χ(G)\leq 2ω(G)-1$ if $G$ is $(P_7, C_4,$ gem)-free. These conclusions generalize some results of Choudum {\em et al} and Lan {\em et al}.

math.CO↗

arXiv2023

Flexoelectricity-stabilized ferroelectric phase with enhanced reliability in ultrathin La:HfO2 films

Doped HfO2 thin films exhibit robust ferroelectric properties even for nanometric thicknesses, are compatible with current Si technology and thus have great potential for the revival of integrated ferroelectrics. Phase control and reliability are core issues for their applications. Here we show that, in (111)-oriented 5%La:HfO2 (HLO) epitaxial thin films deposited on (La0.3Sr0.7)(Al0.65Ta0.35)O3 substrates, the flexoelectric effect, arising from the strain gradient along the films normal, induces a rhombohedral distortion in the otherwise Pca21 orthorhombic structure. Density functional calculations reveal that the distorted structure is indeed more stable than the pure Pca21 structure, when applying an electric field mimicking the flexoelectric field. This rhombohedral distortion greatly improves the fatigue endurance of HLO thin films by further stabilizing the metastable ferroelectric phase against the transition to the thermodynamically stable non-polar monoclinic phase during repetitive cycling. Our results demonstrate that the flexoelectric effect, though negligibly weak in bulk, is crucial to optimize the structure and properties of doped HfO2 thin films with nanometric thicknesses for integrated ferroelectric applications.

cond-mat.mtrl-sci↗

arXiv2023

Adaptive Aggregation for Safety-Critical Control

Safety has been recognized as the central obstacle to preventing the use of reinforcement learning (RL) for real-world applications. Different methods have been developed to deal with safety concerns in RL. However, learning reliable RL-based solutions usually require a large number of interactions with the environment. Likewise, how to improve the learning efficiency, specifically, how to utilize transfer learning for safe reinforcement learning, has not been well studied. In this work, we propose an adaptive aggregation framework for safety-critical control. Our method comprises two key techniques: 1) we learn to transfer the safety knowledge by aggregating the multiple source tasks and a target task through the attention network; 2) we separate the goal of improving task performance and reducing constraint violations by utilizing a safeguard. Experiment results demonstrate that our algorithm can achieve fewer safety violations while showing better data efficiency compared with several baselines.

cs.LG↗

arXiv2023

Self-Supervised Transformer Architecture for Change Detection in Radio Access Networks

Radio Access Networks (RANs) for telecommunications represent large agglomerations of interconnected hardware consisting of hundreds of thousands of transmitting devices (cells). Such networks undergo frequent and often heterogeneous changes caused by network operators, who are seeking to tune their system parameters for optimal performance. The effects of such changes are challenging to predict and will become even more so with the adoption of 5G/6G networks. Therefore, RAN monitoring is vital for network operators. We propose a self-supervised learning framework that leverages self-attention and self-distillation for this task. It works by detecting changes in Performance Measurement data, a collection of time-varying metrics which reflect a set of diverse measurements of the network performance at the cell level. Experimental results show that our approach outperforms the state of the art by 4% on a real-world based dataset consisting of about hundred thousands timeseries. It also has the merits of being scalable and generalizable. This allows it to provide deep insight into the specifics of mode of operation changes while relying minimally on expert knowledge.

cs.LG↗

arXiv2023

Task Placement and Resource Allocation for Edge Machine Learning: A GNN-based Multi-Agent Reinforcement Learning Paradigm

Machine learning (ML) tasks are one of the major workloads in today's edge computing networks. Existing edge-cloud schedulers allocate the requested amounts of resources to each task, falling short of best utilizing the limited edge resources for ML tasks. This paper proposes TapFinger, a distributed scheduler for edge clusters that minimizes the total completion time of ML tasks through co-optimizing task placement and fine-grained multi-resource allocation. To learn the tasks' uncertain resource sensitivity and enable distributed scheduling, we adopt multi-agent reinforcement learning (MARL) and propose several techniques to make it efficient, including a heterogeneous graph attention network as the MARL backbone, a tailored task selection phase in the actor network, and the integration of Bayes' theorem and masking schemes. We first implement a single-task scheduling version, which schedules at most one task each time. Then we generalize to the multi-task scheduling case, in which a sequence of tasks is scheduled simultaneously. Our design can mitigate the expanded decision space and yield fast convergence to optimal scheduling solutions. Extensive experiments using synthetic and test-bed ML task traces show that TapFinger can achieve up to 54.9% reduction in the average task completion time and improve resource efficiency as compared to state-of-the-art schedulers.

cs.MA↗

arXiv2023

Design Considerations of a Coordinative Demand Charge Mitigation Strategy

This paper presents a coordinative demand charge mitigation (DCM) strategy for reducing electricity consumption during system peak periods. Available DCM resources include batteries, diesel generators, controllable loads, and conservation voltage reduction. All resources are directly controlled by load serving entities. A mixed integer linear programming based energy management algorithm is developed to optimally coordinate of DCM resources considering the load payback effect. To better capture system peak periods, two different kinds of load forecast are used: the day-ahead load forecast and the peak-hour probability forecast. Five DCM strategies are compared for reconciling the discrepancy between the two forecasting results. The DCM strategies are tested using actual utility data. Simulation results show that the proposed algorithm can effectively mitigate the demand charge while preventing the system peak from being shifted to the payback hours. We also identify the diminishing return effect, which can help load serving entities optimize the size of their DCM resources.

eess.SY↗

arXiv2023

Energy Storage State-of-Charge Market Model

This paper introduces and rationalizes a new model for bidding and clearing energy storage resources in wholesale energy markets. Charge and discharge bids in this model depend on the storage state-of-charge (SoC). In this setting, storage participants submit different bids for each SoC segment. The system operator monitors the storage SoC and updates their bids accordingly in market clearings. Combined with an optimal bidding design algorithm using dynamic programming, our paper shows that the SoC segment market model provides more accurate representations of the opportunity costs of energy storage compared to existing power-based bidding models. The new model also captures the inherent SoC-dependent operational characteristics of energy storage. We benchmark the SoC segment market model against an existing single-segment model in price-taker and price-influencer simulations. The simulation results show that compared to the existing power-based bidding model, the proposed model improves profits by 10-56% in the price-taker case study; the model also improves total system cost reduction from storage by around 5%, and helps reduce price volatilities in the price-influencer case study.

eess.SY↗

arXiv2023

TrimTail: Low-Latency Streaming ASR with Simple but Effective Spectrogram-Level Length Penalty

In this paper, we present TrimTail, a simple but effective emission regularization method to improve the latency of streaming ASR models. The core idea of TrimTail is to apply length penalty (i.e., by trimming trailing frames, see Fig. 1-(b)) directly on the spectrogram of input utterances, which does not require any alignment. We demonstrate that TrimTail is computationally cheap and can be applied online and optimized with any training loss or any model architecture on any dataset without any extra effort by applying it on various end-to-end streaming ASR networks either trained with CTC loss [1] or Transducer loss [2]. We achieve 100 $\sim$ 200ms latency reduction with equal or even better accuracy on both Aishell-1 and Librispeech. Moreover, by using TrimTail, we can achieve a 400ms algorithmic improvement of User Sensitive Delay (USD) with an accuracy loss of less than 0.2.

cs.SD↗

arXiv2023

Topological classes of rotating black holes

In this paper, we investigate the topological numbers for singly rotating Kerr black holes in arbitrary dimensions and four-dimensional Kerr-Newman black hole. We show that for uncharged black holes, the rotation parameter has a significant effect on the topological number, and for rotating black holes, the dimension of spacetime has a remarkable effect on the topological number too. In addition, we find that the topological numbers of the four-dimensional Kerr and Kerr-Newman black holes are the same, which seems to indicate that the electric charge parameter has no effect on the topological number of rotating black holes. Our current research provides more evidence that the conjecture put forward in Wei et al. [Phys. Rev. Lett. 129, 191101 (2022)], according to which all black hole solutions should be separated into three different topological classes, is accurate, at least in the pure Einstein-Maxwell gravity theory.

gr-qc↗

arXiv2023

Gap Minimization for Knowledge Sharing and Transfer

Learning from multiple related tasks by knowledge sharing and transfer has become increasingly relevant over the last two decades. In order to successfully transfer information from one task to another, it is critical to understand the similarities and differences between the domains. In this paper, we introduce the notion of \emph{performance gap}, an intuitive and novel measure of the distance between learning tasks. Unlike existing measures which are used as tools to bound the difference of expected risks between tasks (e.g., $\mathcal{H}$-divergence or discrepancy distance), we theoretically show that the performance gap can be viewed as a data- and algorithm-dependent regularizer, which controls the model complexity and leads to finer guarantees. More importantly, it also provides new insights and motivates a novel principle for designing strategies for knowledge sharing and transfer: gap minimization. We instantiate this principle with two algorithms: 1. gapBoost, a novel and principled boosting algorithm that explicitly minimizes the performance gap between source and target domains for transfer learning; and 2. gapMTNN, a representation learning algorithm that reformulates gap minimization as semantic conditional matching for multitask learning. Our extensive evaluation on both transfer learning and multitask learning benchmark data sets shows that our methods outperform existing baselines.

cs.LG↗

arXiv2023

A Novel Modular, Reconfigurable Battery Energy Storage System: Design, Control, and Experimentation

This paper presents a novel modular, reconfigurable battery energy storage system. The proposed design is characterized by a tight integration of reconfigurable power switches and DC/DC converters. This characteristic enables isolation of faulty cells from the system and allows fine power control for individual cells toward optimal system-level performance. An optimal power management approach is developed to extensively exploit the merits of the proposed design. Based on receding-horizon convex optimization, this approach aims to minimize the total power losses in charging/discharging while allocating the power in line with each cell's condition to achieve state-of-charge (SoC) and temperature balancing. By appropriate design, the approach manages to regulate the power of a cell across its full SoC range and guarantees the feasibility of the optimization problem. We perform extensive simulations and further develop a lab-scale prototype to validate the proposed system design and power management approach.

eess.SY↗

arXiv2022

Original or Translated? On the Use of Parallel Data for Translation Quality Estimation

Machine Translation Quality Estimation (QE) is the task of evaluating translation output in the absence of human-written references. Due to the scarcity of human-labeled QE data, previous works attempted to utilize the abundant unlabeled parallel corpora to produce additional training data with pseudo labels. In this paper, we demonstrate a significant gap between parallel data and real QE data: for QE data, it is strictly guaranteed that the source side is original texts and the target side is translated (namely translationese). However, for parallel data, it is indiscriminate and the translationese may occur on either source or target side. We compare the impact of parallel data with different translation directions in QE data augmentation, and find that using the source-original part of parallel corpus consistently outperforms its target-original counterpart. Moreover, since the WMT corpus lacks direction information for each parallel sentence, we train a classifier to distinguish source- and target-original bitext, and carry out an analysis of their difference in both style and domain. Together, these findings suggest using source-original parallel data for QE data augmentation, which brings a relative improvement of up to 4.0% and 6.4% compared to undifferentiated data on sentence- and word-level QE tasks respectively.

cs.CL↗

arXiv2022

Multi-Metric AutoRec for High Dimensional and Sparse User Behavior Data Prediction

User behavior data produced during interaction with massive items in the significant data era are generally heterogeneous and sparse, leaving the recommender system (RS) a large diversity of underlying patterns to excavate. Deep neural network-based models have reached the state-of-the-art benchmark of the RS owing to their fitting capabilities. However, prior works mainly focus on designing an intricate architecture with fixed loss function and regulation. These single-metric models provide limited performance when facing heterogeneous and sparse user behavior data. Motivated by this finding, we propose a multi-metric AutoRec (MMA) based on the representative AutoRec. The idea of the proposed MMA is mainly two-fold: 1) apply different $L_p$-norm on loss function and regularization to form different variant models in different metric spaces, and 2) aggregate these variant models. Thus, the proposed MMA enjoys the multi-metric orientation from a set of dispersed metric spaces, achieving a comprehensive representation of user data. Theoretical studies proved that the proposed MMA could attain performance improvement. The extensive experiment on five real-world datasets proves that MMA can outperform seven other state-of-the-art models in predicting unobserved user behavior data.

cs.IR↗