arXiv subjects

Zhi Yu

Publications and source records attributed to Zhi Yu.

At least 37 records · Page 2Linked to original sources

arXiv2024

LORE++: Logical Location Regression Network for Table Structure Recognition with Pre-training

Table structure recognition (TSR) aims at extracting tables in images into machine-understandable formats. Recent methods solve this problem by predicting the adjacency relations of detected cell boxes or learning to directly generate the corresponding markup sequences from the table images. However, existing approaches either count on additional heuristic rules to recover the table structures, or face challenges in capturing long-range dependencies within tables, resulting in increased complexity. In this paper, we propose an alternative paradigm. We model TSR as a logical location regression problem and propose a new TSR framework called LORE, standing for LOgical location REgression network, which for the first time regresses logical location as well as spatial location of table cells in a unified network. Our proposed LORE is conceptually simpler, easier to train, and more accurate than other paradigms of TSR. Moreover, inspired by the persuasive success of pre-trained models on a number of computer vision and natural language processing tasks, we propose two pre-training tasks to enrich the spatial and logical representations at the feature level of LORE, resulting in an upgraded version called LORE++. The incorporation of pre-training in LORE++ has proven to enjoy significant advantages, leading to a substantial enhancement in terms of accuracy, generalization, and few-shot capability compared to its predecessor. Experiments on standard benchmarks against methods of previous paradigms demonstrate the superiority of LORE++, which highlights the potential and promising prospect of the logical location regression paradigm for TSR.

cs.CV↗

arXiv2023

Multi-View Fusion and Distillation for Subgrade Distresses Detection based on 3D-GPR

The application of 3D ground-penetrating radar (3D-GPR) for subgrade distress detection has gained widespread popularity. To enhance the efficiency and accuracy of detection, pioneering studies have attempted to adopt automatic detection techniques, particularly deep learning. However, existing works typically rely on traditional 1D A-scan, 2D B-scan or 3D C-scan data of the GPR, resulting in either insufficient spatial information or high computational complexity. To address these challenges, we introduce a novel methodology for the subgrade distress detection task by leveraging the multi-view information from 3D-GPR data. Moreover, we construct a real multi-view image dataset derived from the original 3D-GPR data for the detection task, which provides richer spatial information compared to A-scan and B-scan data, while reducing computational complexity compared to C-scan data. Subsequently, we develop a novel \textbf{M}ulti-\textbf{V}iew \textbf{V}usion and \textbf{D}istillation framework, \textbf{GPR-MVFD}, specifically designed to optimally utilize the multi-view GPR dataset. This framework ingeniously incorporates multi-view distillation and attention-based fusion to facilitate significant feature extraction for subgrade distresses. In addition, a self-adaptive learning mechanism is adopted to stabilize the model training and prevent performance degeneration in each branch. Extensive experiments conducted on this new GPR benchmark demonstrate the effectiveness and efficiency of our proposed framework. Our framework outperforms not only the existing GPR baselines, but also the state-of-the-art methods in the fields of multi-view learning, multi-modal learning, and knowledge distillation. We will release the constructed multi-view GPR dataset with expert-annotated labels and the source codes of the proposed framework.

cs.CV↗

arXiv2023

Dynamic Data-Free Knowledge Distillation by Easy-to-Hard Learning Strategy

Data-free knowledge distillation (DFKD) is a widely-used strategy for Knowledge Distillation (KD) whose training data is not available. It trains a lightweight student model with the aid of a large pretrained teacher model without any access to training data. However, existing DFKD methods suffer from inadequate and unstable training process, as they do not adjust the generation target dynamically based on the status of the student model during learning. To address this limitation, we propose a novel DFKD method called CuDFKD. It teaches students by a dynamic strategy that gradually generates easy-to-hard pseudo samples, mirroring how humans learn. Besides, CuDFKD adapts the generation target dynamically according to the status of student model. Moreover, We provide a theoretical analysis of the majorization minimization (MM) algorithm and explain the convergence of CuDFKD. To measure the robustness and fidelity of DFKD methods, we propose two more metrics, and experiments shows CuDFKD has comparable performance to state-of-the-art (SOTA) DFKD methods on all datasets. Experiments also present that our CuDFKD has the fastest convergence and best robustness over other SOTA DFKD methods.

cs.CV↗

arXiv2023

Translate the Beauty in Songs: Jointly Learning to Align Melody and Translate Lyrics

Song translation requires both translation of lyrics and alignment of music notes so that the resulting verse can be sung to the accompanying melody, which is a challenging problem that has attracted some interests in different aspects of the translation process. In this paper, we propose Lyrics-Melody Translation with Adaptive Grouping (LTAG), a holistic solution to automatic song translation by jointly modeling lyrics translation and lyrics-melody alignment. It is a novel encoder-decoder framework that can simultaneously translate the source lyrics and determine the number of aligned notes at each decoding step through an adaptive note grouping module. To address data scarcity, we commissioned a small amount of training data annotated specifically for this task and used large amounts of augmented data through back-translation. Experiments conducted on an English-Chinese song translation data set show the effectiveness of our model in both automatic and human evaluation.

cs.CL↗

arXiv2023

LORE: Logical Location Regression Network for Table Structure Recognition

cs.CV↗

arXiv2023

City-scale synthetic individual-level vehicle trip data

Trip data that records each vehicle's trip activity on the road network describes the operation of urban traffic from the individual perspective, and it is extremely valuable for transportation research. However, restricted by data privacy, the trip data of individual-level cannot be opened for all researchers, while the need for it is very urgent. In this paper, we produce a city-scale synthetic individual-level vehicle trip dataset by generating for each individual based on the historical trip data, where the availability and trip data privacy protection are balanced. Privacy protection inevitably affects the availability of data. Therefore, we have conducted numerous experiments to demonstrate the performance and reliability of the synthetic data in different dimensions and at different granularities to help users properly judge the tasks it can perform. The result shows that the synthetic data is consistent with the real data (i.e., historical data) on the aggregated level and reasonable from the individual perspective.

cs.CY↗

arXiv2022

On the derivation of guiding center dynamics without coordinate dependence

The fundament of the classical guiding center theory is gyro-phase averaging, which cannot be well defined over a non-trivial magnetic field topology. The local gyro-phase coordinate frame hides the geometric nature of gyro-symmetry. A coordinate-free geometric representation should be a more appropriate alternative for a deeper understanding of the guiding center dynamics. In this paper, the motion of a charged particle is described by a Lagrangian one-form on a seven-dimensional phase space. The Lagrangian one-form is geometrically decomposed by constructing a coordinate-free gyro-averaging method. As a result, we obtain the coordinate-free expression of the non-relativistic guiding-center dynamics in the time-dependent slow-varying electromagnetic field.

physics.plasm-ph↗

arXiv2022

A data management system for machine learning research of tokamak

In recent years, machine learning (ML) research methods have received increasing attention in the tokamak community. The conventional database (i.e., MDSplus for tokamak) of experimental data has been designed for small group consumption and is mainly aimed at simultaneous visualization of a small amount of data. The ML data access patterns fundamentally differ from traditional data access patterns. The typical MDSplus database is increasingly showing its limitations. We developed a new data management system suitable for tokamak machine learning research based on Experimental Advanced Superconducting Tokamak (EAST) data. The data management system is based on MongoDB and Hierarchical Data Format version 5 (HDF5). Currently, the entire data management has more than 3000 channels of data. The system can provide highly reliable concurrent access. The system includes error correction, MDSplus original data conversion, and high-performance sequence data output. Further, some valuable functions are implemented to accelerate ML model training of fusion, such as bucketing generator, the concatenating buffer, and distributed sequence generation. This data management system is more suitable for fusion machine learning model R\&D than MDSplus, but it can not replace the MDSplus database. The MDSplus database is still the backend for EAST tokamak data acquisition and storage.

physics.plasm-ph↗

arXiv2022

Wireless Communication Using Metal Reflectors: Reflection Modelling and Experimental Verification

Wireless communication using fully passive metal reflectors is a promising technique for coverage expansion, signal enhancement, rank improvement and blind-zone compensation, thanks to its appealing features including zero energy consumption, ultra low cost, signaling- and maintenance-free, easy deployment and full compatibility with existing and future wireless systems. However, a prevalent understanding for reflection by metal plates is based on Snell's Law, i.e., signal can only be received when the observation angle equals to the incident angle, which is valid only when the electrical dimension of the metal plate is extremely large. In this paper, we rigorously derive a general reflection model that is applicable to metal reflectors of any size, any orientation, and any linear polarization. The derived model is given compactly in terms of the radar cross section (RCS) of the metal plate, as a function of its physical dimensions and orientation vectors, as well as the wave polarization and the wave deflection vector, i.e., the change of direction from the incident wave direction to the observation direction. Furthermore, experimental results based on actual field measurements are provided to validate the accuracy of our developed model and demonstrate the great potential of communications using metal reflectors.

cs.IT↗

arXiv2022

EAST discharge prediction without integrating simulation results

In this work, a purely data-driven discharge prediction model was developed and tested without integrating any data or results from simulations. The model was developed based on the experimental data from the Experimental Advanced Superconducting Tokamak (EAST) campaign 2010-2020 discharges and can predict the actual plasma current $I_{p}$, normalized beta $β_{n}$, toroidal beta $β_{t}$, beta poloidal $β_{p}$, electron density $n_{e}$, store energy $W_{mhd}$, loop voltage $V_{loop}$, elongation at plasma boundary $κ$, internal inductance $l_{i}$, q at magnetic axis $q_{0}$, and q at 95% flux surface $q_{95}$. The average similarities of all the selected key diagnostic signals between prediction results and the experimental data are greater than 90%, except for the $V_{loop}$ and $q_{0}$. Before a tokamak experiment, the values of actuator signals are set in the discharge proposal stage, with the model allowing to check the consistency of expected diagnostic signals. The model can give the estimated values of the diagnostic signals to check the reasonableness of the tokamak experimental proposal.

physics.plasm-ph↗

arXiv2022

A machine-learning-based tool for last closed-flux surface reconstruction on tokamaks

Nuclear fusion represents one of the best alternatives for a sustainable source of clean energy. Tokamaks allow to confine fusion plasma with magnetic fields and one of the main challenges in the control of the magnetic configuration is the prediction/reconstruction of the Last Closed-Flux Surface (LCFS). The evolution in time of the LCFS is determined by the interaction of the actuator coils and the internal tokamak plasma. This task requires real-time capable tools able to deal with high-dimensional data as well as with high resolution in time, where the interaction between a wide range of input actuator coils with internal plasma state responses add additional layer of complexity. In this work, we present the application of a novel state of the art machine learning model to the LCFS reconstruction in the Experimental Advanced Superconducting Tokamak (EAST) that learns automatically from the experimental data of EAST. This architecture allows not only offline simulation and testing of a particular control strategy, but can also be embedded in the real-time control system for online magnetic equilibrium reconstruction and prediction. In the real-time modeling test, our approach achieves very high accuracies, with over 99% average similarity in LCFS reconstruction of the entire discharge process.

physics.plasm-ph↗

arXiv2022

City-Scale Holographic Traffic Flow Data based on Vehicular Trajectory Resampling

Despite abundant accessible traffic data, researches on traffic flow estimation and optimization still face the dilemma of detailedness and integrity in the measurement. A dataset of city-scale vehicular continuous trajectories featuring the finest resolution and integrity, as known as the holographic traffic data, would be a breakthrough, for it could reproduce every detail of the traffic flow evolution and reveal the personal mobility pattern within the city. Due to the high coverage of Automatic Vehicle Identification (AVI) devices in Xuancheng city, we constructed one-month continuous trajectories of daily 80,000 vehicles in the city with accurate intersection passing time and no travel path estimation bias. With such holographic traffic data, it is possible to reproduce every detail of the traffic flow evolution. We presented a set of traffic flow data based on the holographic trajectories resampling, covering the whole 482 road segments in the city round the clock, including stationary average speed and flow data of 5-minute intervals and dynamic floating car data.

cs.CY↗

arXiv2021

SentiPrompt: Sentiment Knowledge Enhanced Prompt-Tuning for Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis (ABSA) is an emerging fine-grained sentiment analysis task that aims to extract aspects, classify corresponding sentiment polarities and find opinions as the causes of sentiment. The latest research tends to solve the ABSA task in a unified way with end-to-end frameworks. Yet, these frameworks get fine-tuned from downstream tasks without any task-adaptive modification. Specifically, they do not use task-related knowledge well or explicitly model relations between aspect and opinion terms, hindering them from better performance. In this paper, we propose SentiPrompt to use sentiment knowledge enhanced prompts to tune the language model in the unified framework. We inject sentiment knowledge regarding aspects, opinions, and polarities into prompt and explicitly model term relations via constructing consistency and polarity judgment templates from the ground truth triplets. Experimental results demonstrate that our approach can outperform strong baselines on Triplet Extraction, Pair Extraction, and Aspect Term Extraction with Sentiment Classification by a notable margin.

cs.CL↗

arXiv2021

Cross-modal Image Retrieval with Deep Mutual Information Maximization

In this paper, we study the cross-modal image retrieval, where the inputs contain a source image plus some text that describes certain modifications to this image and the desired image. Prior work usually uses a three-stage strategy to tackle this task: 1) extract the features of the inputs; 2) fuse the feature of the source image and its modified text to obtain fusion feature; 3) learn a similarity metric between the desired image and the source image + modified text by using deep metric learning. Since classical image/text encoders can learn the useful representation and common pair-based loss functions of distance metric learning are enough for cross-modal retrieval, people usually improve retrieval accuracy by designing new fusion networks. However, these methods do not successfully handle the modality gap caused by the inconsistent distribution and representation of the features of different modalities, which greatly influences the feature fusion and similarity learning. To alleviate this problem, we adopt the contrastive self-supervised learning method Deep InforMax (DIM) to our approach to bridge this gap by enhancing the dependence between the text, the image, and their fusion. Specifically, our method narrows the modality gap between the text modality and the image modality by maximizing mutual information between their not exactly semantically identical representation. Moreover, we seek an effective common subspace for the semantically same fusion feature and desired image's feature by utilizing Deep InforMax between the low-level layer of the image encoder and the high-level layer of the fusion network. Extensive experiments on three large-scale benchmark datasets show that we have bridged the modality gap between different modalities and achieve state-of-the-art retrieval performance.

cs.CV↗

arXiv2020

Experiment data-driven modeling of tokamak discharge in EAST

A model for tokamak discharge through deep learning has been done on a superconducting long-pulse tokamak (EAST). This model can use the control signals (i.e. Neutral Beam Injection (NBI), Ion Cyclotron Resonance Heating (ICRH), etc) to model normal discharge without the need for doing real experiments. By using the data-driven methodology, we exploit the temporal sequence of control signals for a large set of EAST discharges to develop a deep learning model for modeling discharge diagnostic signals, such as electron density $n_{e}$, store energy $W_{mhd}$ and loop voltage $V_{loop}$. Comparing the similar methodology, we use Machine Learning techniques to develop the data-driven model for discharge modeling rather than disruption prediction. Up to 95% similarity was achieved for $W_{mhd}$. The first try showed promising results for modeling of tokamak discharge by using the data-driven methodology. The data-driven methodology provides an alternative to physical-driven modeling for tokamak discharge modeling.

physics.plasm-ph↗

arXiv2020

Adaptive-Step Graph Meta-Learner for Few-Shot Graph Classification

Graph classification aims to extract accurate information from graph-structured data for classification and is becoming more and more important in graph learning community. Although Graph Neural Networks (GNNs) have been successfully applied to graph classification tasks, most of them overlook the scarcity of labeled graph data in many applications. For example, in bioinformatics, obtaining protein graph labels usually needs laborious experiments. Recently, few-shot learning has been explored to alleviate this problem with only given a few labeled graph samples of test classes. The shared sub-structures between training classes and test classes are essential in few-shot graph classification. Exiting methods assume that the test classes belong to the same set of super-classes clustered from training classes. However, according to our observations, the label spaces of training classes and test classes usually do not overlap in real-world scenario. As a result, the existing methods don't well capture the local structures of unseen test classes. To overcome the limitation, in this paper, we propose a direct method to capture the sub-structures with well initialized meta-learner within a few adaptation steps. More specifically, (1) we propose a novel framework consisting of a graph meta-learner, which uses GNNs based modules for fast adaptation on graph data, and a step controller for the robustness and generalization of meta-learner; (2) we provide quantitative analysis for the framework and give a graph-dependent upper bound of the generalization error based on our framework; (3) the extensive experiments on real-world datasets demonstrate that our framework gets state-of-the-art results on several few-shot graph classification tasks compared to baselines.

cs.LG↗

arXiv2020

Matching Text with Deep Mutual Information Estimation

Text matching is a core natural language processing research problem. How to retain sufficient information on both content and structure information is one important challenge. In this paper, we present a neural approach for general-purpose text matching with deep mutual information estimation incorporated. Our approach, Text matching with Deep Info Max (TIM), is integrated with a procedure of unsupervised learning of representations by maximizing the mutual information between text matching neural network's input and output. We use both global and local mutual information to learn text representations. We evaluate our text matching approach on several tasks including natural language inference, paraphrase identification, and answer selection. Compared to the state-of-the-art approaches, the experiments show that our method integrated with mutual information estimation learns better text representation and achieves better experimental results of text matching tasks without exploiting pretraining on external data.

cs.CL↗

arXiv2019

Hierarchical Graph Pooling with Structure Learning

Graph Neural Networks (GNNs), which generalize deep neural networks to graph-structured data, have drawn considerable attention and achieved state-of-the-art performance in numerous graph related tasks. However, existing GNN models mainly focus on designing graph convolution operations. The graph pooling (or downsampling) operations, that play an important role in learning hierarchical representations, are usually overlooked. In this paper, we propose a novel graph pooling operator, called Hierarchical Graph Pooling with Structure Learning (HGP-SL), which can be integrated into various graph neural network architectures. HGP-SL incorporates graph pooling and structure learning into a unified module to generate hierarchical representations of graphs. More specifically, the graph pooling operation adaptively selects a subset of nodes to form an induced subgraph for the subsequent layers. To preserve the integrity of graph's topological information, we further introduce a structure learning mechanism to learn a refined graph structure for the pooled graph at each layer. By combining HGP-SL operator with graph neural networks, we perform graph level representation learning with focus on graph classification task. Experimental results on six widely used benchmarks demonstrate the effectiveness of our proposed model.

cs.LG↗