arXiv subjects

Yu Zhou

Publications and source records attributed to Yu Zhou.

At least 289 records · Page 16Linked to original sources

arXiv2023

E2TIMT: Efficient and Effective Modal Adapter for Text Image Machine Translation

Text image machine translation (TIMT) aims to translate texts embedded in images from one source language to another target language. Existing methods, both two-stage cascade and one-stage end-to-end architectures, suffer from different issues. The cascade models can benefit from the large-scale optical character recognition (OCR) and MT datasets but the two-stage architecture is redundant. The end-to-end models are efficient but suffer from training data deficiency. To this end, in our paper, we propose an end-to-end TIMT model fully making use of the knowledge from existing OCR and MT datasets to pursue both an effective and efficient framework. More specifically, we build a novel modal adapter effectively bridging the OCR encoder and MT decoder. End-to-end TIMT loss and cross-modal contrastive loss are utilized jointly to align the feature distribution of the OCR and MT tasks. Extensive experiments show that the proposed method outperforms the existing two-stage cascade models and one-stage end-to-end models with a lighter and faster architecture. Furthermore, the ablation studies verify the generalization of our method, where the proposed modal adapter is effective to bridge various OCR and MT models.

cs.CL↗

arXiv2023

Multi-Teacher Knowledge Distillation For Text Image Machine Translation

Text image machine translation (TIMT) has been widely used in various real-world applications, which translates source language texts in images into another target language sentence. Existing methods on TIMT are mainly divided into two categories: the recognition-then-translation pipeline model and the end-to-end model. However, how to transfer knowledge from the pipeline model into the end-to-end model remains an unsolved problem. In this paper, we propose a novel Multi-Teacher Knowledge Distillation (MTKD) method to effectively distillate knowledge into the end-to-end TIMT model from the pipeline model. Specifically, three teachers are utilized to improve the performance of the end-to-end TIMT model. The image encoder in the end-to-end TIMT model is optimized with the knowledge distillation guidance from the recognition teacher encoder, while the sequential encoder and decoder are improved by transferring knowledge from the translation sequential and decoder teacher models. Furthermore, both token and sentence-level knowledge distillations are incorporated to better boost the translation performance. Extensive experimental results show that our proposed MTKD effectively improves the text image translation performance and outperforms existing end-to-end and pipeline models with fewer parameters and less decoding time, illustrating that MTKD can take advantage of both pipeline and end-to-end models.

cs.CL↗

arXiv2023

A Lightweight Recurrent Learning Network for Sustainable Compressed Sensing

Recently, deep learning-based compressed sensing (CS) has achieved great success in reducing the sampling and computational cost of sensing systems and improving the reconstruction quality. These approaches, however, largely overlook the issue of the computational cost; they rely on complex structures and task-specific operator designs, resulting in extensive storage and high energy consumption in CS imaging systems. In this paper, we propose a lightweight but effective deep neural network based on recurrent learning to achieve a sustainable CS system; it requires a smaller number of parameters but obtains high-quality reconstructions. Specifically, our proposed network consists of an initial reconstruction sub-network and a residual reconstruction sub-network. While the initial reconstruction sub-network has a hierarchical structure to progressively recover the image, reducing the number of parameters, the residual reconstruction sub-network facilitates recurrent residual feature extraction via recurrent learning to perform both feature fusion and deep reconstructions across different scales. In addition, we also demonstrate that, after the initial reconstruction, feature maps with reduced sizes are sufficient to recover the residual information, and thus we achieved a significant reduction in the amount of memory required. Extensive experiments illustrate that our proposed model can achieve a better reconstruction quality than existing state-of-the-art CS algorithms, and it also has a smaller number of network parameters than these algorithms. Our source codes are available at: https://github.com/C66YU/CSRN.

cs.CV↗

arXiv2023

A geometric model for the module category of a skew-gentle algebra

In this article, we realize skew-gentle algebras as skew-tiling algebras associated to admissible partial triangulations of punctured marked surfaces. Based on this, we establish a bijection between tagged permissible curves and certain indecomposable modules, interpret the Auslander-Reiten translation via the tagged rotation, and show intersection-dimension formulas. As an application, we classify support $τ$-tilting modules via maximal collections of non-crossing tagged generalized permissible curves.

math.RT↗

arXiv2023

Dual-distribution discrepancy with self-supervised refinement for anomaly detection in medical images

Medical anomaly detection is a crucial yet challenging task aimed at recognizing abnormal images to assist in diagnosis. Due to the high-cost annotations of abnormal images, most methods utilize only known normal images during training and identify samples deviating from the normal profile as anomalies in the testing phase. Many readily available unlabeled images containing anomalies are thus ignored in the training phase, restricting the performance. To solve this problem, we introduce one-class semi-supervised learning (OC-SSL) to utilize known normal and unlabeled images for training, and propose Dual-distribution Discrepancy for Anomaly Detection (DDAD) based on this setting. Ensembles of reconstruction networks are designed to model the distribution of normal images and the distribution of both normal and unlabeled images, deriving the normative distribution module (NDM) and unknown distribution module (UDM). Subsequently, the intra-discrepancy of NDM and inter-discrepancy between the two modules are designed as anomaly scores. Furthermore, we propose a new perspective on self-supervised learning, which is designed to refine the anomaly scores rather than detect anomalies directly. Five medical datasets, including chest X-rays, brain MRIs and retinal fundus images, are organized as benchmarks for evaluation. Experiments on these benchmarks comprehensively compare a wide range of anomaly detection methods and demonstrate that our method achieves significant gains and outperforms the state-of-the-art. Code and organized benchmarks are available at https://github.com/caiyu6666/DDAD-ASR.

cs.CV↗

arXiv2023

Quaternion Orthogonal Transformer for Facial Expression Recognition in the Wild

Facial expression recognition (FER) is a challenging topic in artificial intelligence. Recently, many researchers have attempted to introduce Vision Transformer (ViT) to the FER task. However, ViT cannot fully utilize emotional features extracted from raw images and requires a lot of computing resources. To overcome these problems, we propose a quaternion orthogonal transformer (QOT) for FER. Firstly, to reduce redundancy among features extracted from pre-trained ResNet-50, we use the orthogonal loss to decompose and compact these features into three sets of orthogonal sub-features. Secondly, three orthogonal sub-features are integrated into a quaternion matrix, which maintains the correlations between different orthogonal components. Finally, we develop a quaternion vision transformer (Q-ViT) for feature classification. The Q-ViT adopts quaternion operations instead of the original operations in ViT, which improves the final accuracies with fewer parameters. Experimental results on three in-the-wild FER datasets show that the proposed QOT outperforms several state-of-the-art models and reduces the computations.

cs.CV↗

arXiv2023

Cross-hospital Sepsis Early Detection via Semi-supervised Optimal Transport with Self-paced Ensemble

Leveraging machine learning techniques for Sepsis early detection and diagnosis has attracted increasing interest in recent years. However, most existing methods require a large amount of labeled training data, which may not be available for a target hospital that deploys a new Sepsis detection system. More seriously, as treated patients are diversified between hospitals, directly applying a model trained on other hospitals may not achieve good performance for the target hospital. To address this issue, we propose a novel semi-supervised transfer learning framework based on optimal transport theory and self-paced ensemble for Sepsis early detection, called SPSSOT, which can efficiently transfer knowledge from the source hospital (with rich labeled data) to the target hospital (with scarce labeled data). Specifically, SPSSOT incorporates a new optimal transport-based semi-supervised domain adaptation component that can effectively exploit all the unlabeled data in the target hospital. Moreover, self-paced ensemble is adapted in SPSSOT to alleviate the class imbalance issue during transfer learning. In a nutshell, SPSSOT is an end-to-end transfer learning method that automatically selects suitable samples from two domains (hospitals) respectively and aligns their feature spaces. Extensive experiments on two open clinical datasets, MIMIC-III and Challenge, demonstrate that SPSSOT outperforms state-of-the-art transfer learning methods by improving 1-3% of AUC.

cs.LG↗

arXiv2023

Overview and Performance Analysis of Various Waveforms in High Mobility Scenarios

In the high-mobility scenarios of next-generation wireless communication systems (beyond 5G/6G), the performance of orthogonal frequency division multiplexing (OFDM) deteriorates drastically due to the loss of orthogonality between the subcarriers caused by large Doppler frequency shifts. Various emerging waveforms have been proposed for fast time-varying channels with excellent results. In this paper, we classify these waveforms from the perspective of their modulation domain and establish a unified framework to provide a comprehensive system structure comparison. Then we analyze bit error rate (BER) performance of each waveform in doubly selective channels. Through the discussions on their complexity and compatibility with OFDM systems, we finally give the candidate waveform suggestions.

eess.SP↗

arXiv2023

Generating time-domain linear cluster state by recycling superconducting qubits

Cluster states, a type of highly entangled state, are essential resources for quantum information processing. Here we demonstrated the generation of a time-domain linear cluster state (t-LCS) using a superconducting quantum circuit consisting of only two transmon qubits. By recycling the physical qubits, the t-LCS equivalent up to four physical qubits was validated by quantum state tomography with fidelity of 59%. We further confirmed the true generation of t-LCS by examining the expectation value of an entanglement witness. Our demonstrated protocol of t-LCS generation allows efficient use of physical qubits which could lead to resource-efficient execution of quantum circuits on large scale.

quant-ph↗

arXiv2023

On Hill's Worst-Case Guarantee for Indivisible Bads

When allocating objects among agents with equal rights, people often evaluate the fairness of an allocation rule by comparing their received utilities to a benchmark share - a function only of her own valuation and the number of agents. This share is called a guarantee if for any profile of valuations there is an allocation ensuring the share of every agent. When the objects are indivisible goods, Budish [J. Political Econ., 2011] proposed MaxMinShare, i.e., the least utility of a bundle in the best partition of the objects, which is unfortunately not a guarantee. Instead, an earlier pioneering work by Hill [Ann. Probab., 1987] proposed for a share the worst-case MaxMinShare over all valuations with the same largest possible single-object value. Although Hill's share is more conservative than the MaxMinShare, it is an actual guarantee and its computation is elementary, unlike that of the MaxMinShare which involves solving an NP-hard problem. We apply Hill's approach to the allocation of indivisible bads (objects with disutilities or costs), and characterise the tight closed form of the worst-case MinMaxShare for a given value of the worst bad. We argue that Hill's share for allocating bads is effective in the sense of being close to the original MinMaxShare value, and there is much to learn about the guarantee an agent can be offered from the disutility of her worst single bad. Furthermore, we prove that the monotonic cover of Hill's share is the best guarantee that can be achieved in Hill's model for all allocation instances.

cs.GT↗

arXiv2023

Machine learning phase transitions of the three-dimensional Ising universality class

Exploration of the QCD phase diagram and critical point is one of the main goals in current relativistic heavy-ion collisions. The QCD critical point is expected to belong to a three-dimensional (3D) Ising universality class. Machine learning techniques are found to be powerful in distinguishing different phases of matter and provide a new way to study the phase diagram. We investigate phase transitions in the 3D cubic Ising model using supervised learning methods. It is found that a 3D convolutional neural network can be trained to effectivelly predict physical quantities in different spin configurations. With a uniform neural network architecture, it can encode phases of matter and identify both second- and first-order phase transitions. The important features that discriminate different phases in the classification processes are investigated. These findings can help study and understand QCD phase transitions in relativistic heavy-ion collisions.

nucl-th↗

arXiv2023

Test Reuse Based on Adaptive Semantic Matching across Android Mobile Applications

Automatic test generation can help verify and develop the behavior of mobile applications. Test reuse based on semantic similarities between applications of the same category has been utilized to reduce the manual effort of Graphical User Interface (GUI) testing. However, most of the existing studies fail to solve the semantic problem of event matching, which leads to the failure of test reuse. To overcome this challenge, we propose TRASM (Test Reuse based on Adaptive Semantic Matching), a test reuse approach based on adaptive strategies to find a better event matching across android mobile applications. TRASM first performs GUI events deduplication on the initial test set obtained from test generation, and then employs an adaptive strategy to find better event matching, which enables reusing the existing test. Preliminary experiments with comparison to baseline methods on 15 applications demonstrate that TRASM can improve the precision of GUI event matching while reducing the failure of test reuse and the running time required for test reuse.

cs.SE↗

arXiv2022

Mutation graph of support $τ$-tilting modules over a skew-gentle algebra

Let $\mathcal{D}$ be a Hom-finite, Krull-Schmidt, 2-Calabi-Yau triangulated category with a rigid object $R$. Let $Λ=\operatorname{End}_{\mathcal{D}}R$ be the endomorphism algebra of $R$. We introduce the notion of mutation of maximal rigid objects in the two-term subcategory $R\ast R[1]$ via exchange triangles, which is shown to be compatible with mutation of support $τ$-tilting $Λ$-modules. In the case that $\mathcal{D}$ is the cluster category arising from a punctured marked surface, it is shown that the graph of mutations of support $τ$-tilting $Λ$-modules is isomorphic to the graph of flips of certain collections of tagged arcs on the surface, which is moreover proved to be connected. As a direct consequence, the mutation graph of support $τ$-tilting modules over a skew-gentle algebra is connected.

math.RT↗

arXiv2022

Life-long Learning for Multilingual Neural Machine Translation with Knowledge Distillation

A common scenario of Multilingual Neural Machine Translation (MNMT) is that each translation task arrives in a sequential manner, and the training data of previous tasks is unavailable. In this scenario, the current methods suffer heavily from catastrophic forgetting (CF). To alleviate the CF, we investigate knowledge distillation based life-long learning methods. Specifically, in one-tomany scenario, we propose a multilingual distillation method to make the new model (student) jointly learn multilingual output from old model (teacher) and new task. In many-to one scenario, we find that direct distillation faces the extreme partial distillation problem, and we propose two different methods to address it: pseudo input distillation and reverse teacher distillation. The experimental results on twelve translation tasks show that the proposed methods can better consolidate the previous knowledge and sharply alleviate the CF.

cs.CL↗

arXiv2022

Renewing Iterative Self-labeling Domain Adaptation with Application to the Spine Motion Prediction

The area of transfer learning comprises supervised machine learning methods that cope with the issue when the training and testing data have different input feature spaces or distributions. In this work, we propose a novel transfer learning algorithm called Renewing Iterative Self-labeling Domain Adaptation (Re-ISDA). In this work, we propose a novel transfer learning algorithm called Renewing Iterative Self-labeling Domain Adaptation (Re-ISDA).

cs.LG↗

arXiv2022

Beyond Instance Discrimination: Relation-aware Contrastive Self-supervised Learning

Contrastive self-supervised learning (CSL) based on instance discrimination typically attracts positive samples while repelling negatives to learn representations with pre-defined binary self-supervision. However, vanilla CSL is inadequate in modeling sophisticated instance relations, limiting the learned model to retain fine semantic structure. On the one hand, samples with the same semantic category are inevitably pushed away as negatives. On the other hand, differences among samples cannot be captured. In this paper, we present relation-aware contrastive self-supervised learning (ReCo) to integrate instance relations, i.e., global distribution relation and local interpolation relation, into the CSL framework in a plug-and-play fashion. Specifically, we align similarity distributions calculated between the positive anchor views and the negatives at the global level to exploit diverse similarity relations among instances. Local-level interpolation consistency between the pixel space and the feature space is applied to quantitatively model the feature differences of samples with distinct apparent similarities. Through explicitly instance relation modeling, our ReCo avoids irrationally pushing away semantically identical samples and carves a well-structured feature space. Extensive experiments conducted on commonly used benchmarks justify that our ReCo consistently gains remarkable performance improvements.

cs.CV↗

arXiv2022

Non-invasive color imaging through scattering medium under broadband illumination

Due to the complex of mixed spectral point spread function within memory effect range, it is unreliable and slow to use speckle correlation technology for non-invasive imaging through scattering medium under broadband illumination. The contrast of the speckles will drastically drop as the light source's spectrum width increases. Here, we propose a method for producing the optical transfer function with several speckle frames within memory effect range to image under broadband illumination. The method can be applied to image amplitude and color objects under white LED illumination. Compared to other approaches of imaging under broadband illumination, such as deep learning and modified phase retrieval, our method can provide more stable results with faster convergence speed, which can be applied in high speed scattering imaging under natural light illumination.

physics.optics↗

arXiv2022

Improving End-to-End Text Image Translation From the Auxiliary Text Translation Task

End-to-end text image translation (TIT), which aims at translating the source language embedded in images to the target language, has attracted intensive attention in recent research. However, data sparsity limits the performance of end-to-end text image translation. Multi-task learning is a non-trivial way to alleviate this problem via exploring knowledge from complementary related tasks. In this paper, we propose a novel text translation enhanced text image translation, which trains the end-to-end model with text translation as an auxiliary task. By sharing model parameters and multi-task training, our model is able to take full advantage of easily-available large-scale text parallel corpus. Extensive experimental results show our proposed method outperforms existing end-to-end methods, and the joint multi-task learning with both text translation and recognition tasks achieves better results, proving translation and recognition auxiliary tasks are complementary.

cs.CL↗