arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 1,009 records · Page 56Linked to original sources

arXiv2023

Integrated Artificial Neural Network with Trainable Activation Function Enabled by Topological Insulator-based Spin-Orbit Torque Devices

Non-volatile memristors offer a salient platform for artificial neural network (ANN), but the integration of different function blocks into one hardware system remains challenging. Here we demonstrate the implementation of brain-like synaptic (SOT-S) and neuronal (SOT-N) functions in the Bi2Te3/CrTe2 heterostructure-based spin-orbit torque (SOT) device. The SOT-S unit exhibits highly linear (linearity error < 4.19%) and symmetrical long-term potentiation/depression process, resulting in better performance compared to other memristor synapses. Meanwhile, the Sigmoid-shape transition curve inherited in the SOT-N cell replaces the software-based activation function block, hence reducing the system complexity. On this basis, we employ a serial-connected, voltage-mode sensing ANN architecture to enhance the vector-matrix multiplication signal strength with low reading error of 0.61%. Furthermore, the trainable activation function of SOT-N enables the integrated SOT-ANN to execute the Batch Normalization algorithm and activation operation within one clock cycle, which bring about improved on/off-chip training performance close to the ideal baseline.

cond-mat.mes-hall↗

arXiv2023

Some nontrivial secondary Adams differentials on the fourth line

Let $p \geq 5$ be an odd prime. Using the correspondence between secondary Adams differentials and secondary algebraic Novikov differentials, we compute four families of nontrivial secondary differentials on the fourth line of the Adams spectral sequence. We also recover all secondary differentials on the first three lines of the Adams spectral sequence.

math.AT↗

arXiv2023

A detection of the layered structure of nearby open clusters

We applied the newly developed rose diagram overlay method to detect the layered structure of 88 nearby open clusters ($\leq$500~pc) on the three projections after the distance correction of their member stars, based on the catalog in literature. The results show that with the rose diagram overlay method, a total of 74 clusters in our sample have a layered structure, while the remaining clusters are without a clear layered structure. We for the first time defined the layered structure parameters for the sample clusters. Meanwhile, we found that the layered circle core area ($s$) has a strong positive correlation with the number of cluster members, while the kernel instability index ($η$) has a strong negative correlation with the number of cluster members. Our study provides a novel perspective for the detection of the layered structure of open clusters.

astro-ph.GA↗

arXiv2023

Temporal Consistent Automatic Video Colorization via Semantic Correspondence

Video colorization task has recently attracted wide attention. Recent methods mainly work on the temporal consistency in adjacent frames or frames with small interval. However, it still faces severe challenge of the inconsistency between frames with large interval.To address this issue, we propose a novel video colorization framework, which combines semantic correspondence into automatic video colorization to keep long-range consistency. Firstly, a reference colorization network is designed to automatically colorize the first frame of each video, obtaining a reference image to supervise the following whole colorization process. Such automatically colorized reference image can not only avoid labor-intensive and time-consuming manual selection, but also enhance the similarity between reference and grayscale images. Afterwards, a semantic correspondence network and an image colorization network are introduced to colorize a series of the remaining frames with the help of the reference. Each frame is supervised by both the reference image and the immediately colorized preceding frame to improve both short-range and long-range temporal consistency. Extensive experiments demonstrate that our method outperforms other methods in maintaining temporal consistency both qualitatively and quantitatively. In the NTIRE 2023 Video Colorization Challenge, our method ranks at the 3rd place in Color Distribution Consistency (CDC) Optimization track.

cs.CV↗

arXiv2023

Tuning Language Models as Training Data Generators for Augmentation-Enhanced Few-Shot Learning

Recent studies have revealed the intriguing few-shot learning ability of pretrained language models (PLMs): They can quickly adapt to a new task when fine-tuned on a small amount of labeled data formulated as prompts, without requiring abundant task-specific annotations. Despite their promising performance, most existing few-shot approaches that only learn from the small training set still underperform fully supervised training by nontrivial margins. In this work, we study few-shot learning with PLMs from a different perspective: We first tune an autoregressive PLM on the few-shot samples and then use it as a generator to synthesize a large amount of novel training samples which augment the original training set. To encourage the generator to produce label-discriminative samples, we train it via weighted maximum likelihood where the weight of each token is automatically adjusted based on a discriminative meta-learning objective. A classification PLM can then be fine-tuned on both the few-shot and the synthetic samples with regularization for better generalization and stability. Our approach FewGen achieves an overall better result across seven classification tasks of the GLUE benchmark than existing few-shot learning methods, improving no-augmentation methods by 5+ average points, and outperforming augmentation methods by 3+ average points.

cs.CL↗

arXiv2023

A Self-Training Framework Based on Multi-Scale Attention Fusion for Weakly Supervised Semantic Segmentation

Weakly supervised semantic segmentation (WSSS) based on image-level labels is challenging since it is hard to obtain complete semantic regions. To address this issue, we propose a self-training method that utilizes fused multi-scale class-aware attention maps. Our observation is that attention maps of different scales contain rich complementary information, especially for large and small objects. Therefore, we collect information from attention maps of different scales and obtain multi-scale attention maps. We then apply denoising and reactivation strategies to enhance the potential regions and reduce noisy areas. Finally, we use the refined attention maps to retrain the network. Experiments showthat our method enables the model to extract rich semantic information from multi-scale images and achieves 72.4% mIou scores on both the PASCAL VOC 2012 validation and test sets. The code is available at https://bupt-ai-cz.github.io/SMAF.

cs.CV↗

arXiv2023

A Unifying Framework of Attention-based Neural Load Forecasting

Accurate load forecasting is critical for reliable and efficient planning and operation of electric power grids. In this paper, we propose a unifying deep learning framework for load forecasting, which includes time-varying feature weighting, hierarchical temporal attention, and feature-reinforced error correction. Our framework adopts a modular design with good generalization capability. First, the feature-weighting mechanism assigns input features with temporal weights. Second, a recurrent encoder-decoder structure with hierarchical attention is developed as a load predictor. The hierarchical attention enables a similar day selection, which re-evaluates the importance of historical information at each time step. Third, we develop an error correction module that explores the errors and learned feature hidden information to further improve the model's forecasting performance. Experimental results demonstrate that our proposed framework outperforms existing methods on two public datasets and performance metrics, with the feature weighting mechanism and error correction module being critical to achieving superior performance. Our framework provides an effective solution to the electric load forecasting problem, which can be further adapted to many other forecasting tasks.

cs.LG↗

arXiv2023

Rotating Bardeen black hole surrounded by perfect fluid dark matter as a particle accelerator

We study the event horizon of a rotating Bardeen black hole surrounded by perfect fluid dark matter and the black hole as a particle accelerator. The black hole is represented by four parameters: mass $M$, rotation parameter $a$, dark matter parameter $α$ and magnetic charge $g$. It is interesting that when we determine the values of magnetic charge $g$ and dark matter parameters $α$ we can get a critical rotation parameter $a_E$ and then we get a contour plane with $Δ$= 0 taking three parameters as coordinates. We also derive the effective potential of the particle and the centre-of-mass (CM) energy of the two particles outside the black hole by using the motion equation of the particle in the equatorial plane of the black hole. We find that the CM energy depends not only on the rotation parameter $a$, but also on the parameters $g$ and $α$. We discuss the CM energy for two particles colliding at the black hole horizon in the extreme and non-extreme cases, respectively. It is found that the CM energy can be arbitrarily high when the angular momentum of one of the two particles is the critical angular momentum under the background of extreme black holes and there is no such result for non-extreme black holes, because the particles do not reach the black hole horizon when the angular momentum of the particles is critical angular momentum. Therefore, we prove that the rotating Bardeen black hole surrounded by perfect fluid dark matter can be used as a particle accelerator.

gr-qc↗

arXiv2023

Silicon photonics-integrated time-domain balanced homodyne detector in continuous-variable quantum key distribution

We designed and experimentally demonstrated a silicon photonics-integrated time-domain balanced homodyne detector (TBHD), whose optical part has dimensions of 1.5 mm * 0.4 mm. To automatically and accurately balance the detector, new variable optical attenuators were used, and a common mode rejection ratio of 86.9 dB could be achieved. In the quantum tomography experiment, the density matrix and Wigner function of a coherent state were reconstructed with 99.97 % fidelity. The feasibility of this TBHD in a continuous-variable quantum key distribution (CVQKD) system was also demonstrated. This facilitates the integration of the optical circuits of the CVQKD system based on the GG02 protocol on the silicon photonics chip using TBHD.

quant-ph↗

arXiv2023

Synergy Between Excluded Volume Effect with Co-embedded Microparticles and Chemical Doping in Carbon Nanotube Network-based Composites to Enhance Thermoelectric Power Factor

There is a growing momentum in recent thermoelectric materials research for flexible materials which enhance power output and efficiency at human wearable temperatures. In our previous work, we established the method to improve thermoelectric properties with co-embedding microparticles in carbon nanotube (CNT) network-based composites. In this work, we investigate the synergy of excluded volume effect by silica microparticles with CNT doping toward the enhancement of thermoelectric power factor. We find that the power factor can be enhanced to 28.46 μW/mK2 when the CNT is doped with 0.5 μg/ml of 1,5,7-triazabicyclo [4.4.0] dec-5-ene (TBD) before co-embedding silica microparticles within the carbon nanotube-based composite. We also fabricated and demonstrated flexible wearable thermoelectric devices made of the developed composites that exhibits a power output of 0.025 μW/cm2 at a temperature differential of only 4 K.

physics.app-ph↗

arXiv2023

On LinDistFlow Model Congestion Pricing: Bounding the Changes in Power Tariffs

The optimal power flow (OPF) problem is an important mathematical program that aims at obtaining the best operating point of an electric power grid. The optimization problem typically minimizes the total generation cost subject to certain physical constraints of the system. The so-called linearized distribution flow (LinDistFlow) model leverages a set of linear equations to approximate the nonlinear AC power flows. In this paper, we consider an OPF problem based on the LinDistFlow model for a single-phase radial power network. We derive closed-form solutions to the marginal values of both real and reactive power demands. We also derive upper bounds on the congestion price (a.k.a. `shadow price'), which denotes the change in marginal demand prices when the apparent power flow limits of certain lines are binding at optimum. Various cases of our result are discussed while simulations are carried out on a $141$-bus radial power network.

math.OC↗

arXiv2023

An Adaptive Policy to Employ Sharpness-Aware Minimization

Sharpness-aware minimization (SAM), which searches for flat minima by min-max optimization, has been shown to be useful in improving model generalization. However, since each SAM update requires computing two gradients, its computational cost and training time are both doubled compared to standard empirical risk minimization (ERM). Recent state-of-the-arts reduce the fraction of SAM updates and thus accelerate SAM by switching between SAM and ERM updates randomly or periodically. In this paper, we design an adaptive policy to employ SAM based on the loss landscape geometry. Two efficient algorithms, AE-SAM and AE-LookSAM, are proposed. We theoretically show that AE-SAM has the same convergence rate as SAM. Experimental results on various datasets and architectures demonstrate the efficiency and effectiveness of the adaptive policy.

cs.LG↗

arXiv2023

Frequency tunable, cavity-enhanced single erbium quantum emitter in the telecom band

Single quantum emitters embedded in solid-state hosts are an ideal platform for realizing quantum information processors and quantum network nodes. Among the currently-investigated candidates, Er$^{3+}$ ions are particularly appealing due to their 1.5 $μ$m optical transition in the telecom band as well as their long spin coherence times. However, the long lifetimes of the excited state -- generally in excess of 1 ms -- along with the inhomogeneous broadening of the optical transition result in significant challenges. Photon emission rates are prohibitively small, and different emitters generally create photons with distinct spectra, thereby preventing multi-photon interference -- a requirement for building large-scale, multi-node quantum networks. Here we solve this challenge by demonstrating for the first time linear Stark tuning of the emission frequency of a single Er$^{3+}$ ion. Our ions are embedded in a lithium niobate crystal and couple evanescently to a silicon nano-photonic crystal cavity that provides an up to 143 increase of the measured decay rate. By applying an electric field along the crystal c-axis, we achieve a Stark tuning greater than the ion's linewidth without changing the single-photon emission statistics of the ion. These results are a key step towards rare earth ion-based quantum networks.

quant-ph↗

arXiv2023

Understanding Shared Speech-Text Representations

Recently, a number of approaches to train speech models by incorpo-rating text into end-to-end models have been developed, with Mae-stro advancing state-of-the-art automatic speech recognition (ASR)and Speech Translation (ST) performance. In this paper, we expandour understanding of the resulting shared speech-text representationswith two types of analyses. First we examine the limits of speech-free domain adaptation, finding that a corpus-specific duration modelfor speech-text alignment is the most important component for learn-ing a shared speech-text representation. Second, we inspect the sim-ilarities between activations of unimodal (speech or text) encodersas compared to the activations of a shared encoder. We find that theshared encoder learns a more compact and overlapping speech-textrepresentation than the uni-modal encoders. We hypothesize that thispartially explains the effectiveness of the Maestro shared speech-textrepresentations.

cs.CL↗

arXiv2023

Detection of Pavement Cracks by Deep Learning Models of Transformer and UNet

Fracture is one of the main failure modes of engineering structures such as buildings and roads. Effective detection of surface cracks is significant for damage evaluation and structure maintenance. In recent years, the emergence and development of deep learning techniques have shown great potential to facilitate surface crack detection. Currently, most reported tasks were performed by a convolutional neural network (CNN), while the limitation of CNN may be improved by the transformer architecture introduced recently. In this study, we investigated nine promising models to evaluate their performance in pavement surface crack detection by model accuracy, computational complexity, and model stability. We created 711 images of 224 by 224 pixels with crack labels, selected an optimal loss function, compared the evaluation metrics of the validation dataset and test dataset, analyzed the data details, and checked the segmentation outcomes of each model. We find that transformer-based models generally are easier to converge during the training process and have higher accuracy, but usually exhibit more memory consumption and low processing efficiency. Among nine models, SwinUNet outperforms the other two transformers and shows the highest accuracy among nine models. The results should shed light on surface crack detection by various deep-learning models and provide a guideline for future applications in this field.

cs.CV↗

arXiv2023

Variable stars detection in the field of open cluster NGC 188

This work presents the charge-coupled device (CCD) photometric survey of the old open cluster NGC 188. Time-series V-band photometric observations were conducted for ten nights in January 2017 using the Nanshan One-meter Wide-field Telescope (NOWT) to search for variable stars in the field of the cluster field. A total of 25 variable stars, including one new variable star, were detected in the target field. Among the detected variables, 16 are cluster member stars, and the others are identified as field stars. The periods, radial velocities, effective temperatures, and classifications of the detected variables are discussed in this work. Most of the stars' effective temperatures are between 4200 K and 6600 K, indicating their spectral types are G or K. The newly discovered variable is probably a W UMa system. In this study, a known cluster variable star (V21 = V0769 Cep) is classified as an EA-type variable star based on the presence of an 0.5 magnitude eclipse in its light curve.

astro-ph.SR↗

arXiv2023

Mastering Asymmetrical Multiplayer Game with Multi-Agent Asymmetric-Evolution Reinforcement Learning

Asymmetrical multiplayer (AMP) game is a popular game genre which involves multiple types of agents competing or collaborating with each other in the game. It is difficult to train powerful agents that can defeat top human players in AMP games by typical self-play training method because of unbalancing characteristics in their asymmetrical environments. We propose asymmetric-evolution training (AET), a novel multi-agent reinforcement learning framework that can train multiple kinds of agents simultaneously in AMP game. We designed adaptive data adjustment (ADA) and environment randomization (ER) to optimize the AET process. We tested our method in a complex AMP game named Tom \& Jerry, and our AIs trained without using any human data can achieve a win rate of 98.5% against top human players over 65 matches. The ablation experiments indicated that the proposed modules are beneficial to the framework.

cs.AI↗

arXiv2023

Electric-field-induced topological changes in multilamellar and in confined lipid membranes

It is well known that lipid membranes respond to a threshold transmembrane electric field through a reversible mechanism called electroporation, where hydrophilic water pores form across the membrane, an effect widely used in biological systems. The effect of such fields on interfacially-confined (stacked or supported) lipid membranes, on the other hand, which may strongly modulate interfacial properties, has not to our knowledge been previously studied. Motivated by recent surface forces experiments showing a striking effect of electric fields on lubrication by confined lipid bilayers, we carried out all-atom molecular dynamics simulations of such membranes under transverse electric fields. We find that in addition to electroporation, a new feature emerges of locally merged bilayers which act to bridge the confining interfaces. These features shed light on the remodelling of confined lipid membrane stacks by electric fields, and provides insight into how such fields may modulate frictional and more generally surface interactions in the presence of lipid-based boundary layers.

cond-mat.soft↗