arXiv subjects

Lin Yao

Publications and source records attributed to Lin Yao.

At least 55 records · Page 3Linked to original sources

arXiv2024

Node-Aligned Graph-to-Graph (NAG2G): Elevating Template-Free Deep Learning Approaches in Single-Step Retrosynthesis

Single-step retrosynthesis (SSR) in organic chemistry is increasingly benefiting from deep learning (DL) techniques in computer-aided synthesis design. While template-free DL models are flexible and promising for retrosynthesis prediction, they often ignore vital 2D molecular information and struggle with atom alignment for node generation, resulting in lower performance compared to the template-based and semi-template-based methods. To address these issues, we introduce Node-Aligned Graph-to-Graph (NAG2G), a transformer-based template-free DL model. NAG2G combines 2D molecular graphs and 3D conformations to retain comprehensive molecular details and incorporates product-reactant atom mapping through node alignment which determines the order of the node-by-node graph outputs process in an auto-regressive manner. Through rigorous benchmarking and detailed case studies, we have demonstrated that NAG2G stands out with its remarkable predictive accuracy on the expansive datasets of USPTO-50k and USPTO-FULL. Moreover, the model's practical utility is underscored by its successful prediction of synthesis pathways for multiple drug candidate molecules. This not only proves NAG2G's robustness but also its potential to revolutionize the prediction of complex chemical synthesis processes for future synthetic route design tasks.

cs.LG↗

arXiv2023

Single-image based deep learning for precise atomic defects identification

Defect engineering has been profoundly employed to confer desirable functionality to materials that pristine lattices inherently lack. Although single atomic-resolution scanning transmission electron microscopy (STEM) images are widely accessible for defect engineering, harnessing atomic-scale images containing various defects through traditional image analysis methods is hindered by random noise and human bias. Yet the rise of deep learning (DL) offering an alternative approach, its widespread application is primarily restricted by the need for large amounts of training data with labeled ground truth. In this study, we propose a two-stage method to address the problems of high annotation cost and image noise in the detection of atomic defects in monolayer 2D materials. In the first stage, to tackle the issue of data scarcity, we employ a two-state transformation network based on U-GAT-IT for adding realistic noise to simulated images with pre-located ground truth labels, thereby infinitely expanding the training dataset. In the second stage, atomic defects in monolayer 2D materials are effectively detected with high accuracy using U-Net models trained with the data generated in the first stage, avoiding random noise and human bias issues. In both stages, we utilize segmented unit-cell-level images to simplify the model's task and enhance its accuracy. Our results demonstrate that not only sulfur vacancies, we are also able to visualize oxygen dopants in monolayer MoS2, which are usually overwhelmed by random background noise. As the training was based on a few segmented unit-cell-level realistic images, this method can be readily extended to other 2D materials. Therefore, our results outline novel ways to train the model with minimized datasets, offering great opportunities to fully exploit the power of machine learning (ML) applicable to a broad materials science community.

cond-mat.mtrl-sci↗

arXiv2023

CCMB: A Large-scale Chinese Cross-modal Benchmark

Vision-language pre-training (VLP) on large-scale datasets has shown premier performance on various downstream tasks. In contrast to plenty of available benchmarks with English corpus, large-scale pre-training datasets and downstream datasets with Chinese corpus remain largely unexplored. In this work, we build a large-scale high-quality Chinese Cross-Modal Benchmark named CCMB for the research community, which contains the currently largest public pre-training dataset Zero and five human-annotated fine-tuning datasets for downstream tasks. Zero contains 250 million images paired with 750 million text descriptions, plus two of the five fine-tuning datasets are also currently the largest ones for Chinese cross-modal downstream tasks. Along with the CCMB, we also develop a VLP framework named R2D2, applying a pre-Ranking + Ranking strategy to learn powerful vision-language representations and a two-way distillation method (i.e., target-guided Distillation and feature-guided Distillation) to further enhance the learning capability. With the Zero and the R2D2 VLP framework, we achieve state-of-the-art performance on twelve downstream datasets from five broad categories of tasks including image-text retrieval, image-text matching, image caption, text-to-image generation, and zero-shot image classification. The datasets, models, and codes are available at https://github.com/yuxie11/R2D2

cs.CV↗

arXiv2023

A Human-Machine Joint Learning Framework to Boost Endogenous BCI Training

Brain-computer interfaces (BCIs) provide a direct pathway from the brain to external devices and have demonstrated great potential for assistive and rehabilitation technologies. Endogenous BCIs based on electroencephalogram (EEG) signals, such as motor imagery (MI) BCIs, can provide some level of control. However, mastering spontaneous BCI control requires the users to generate discriminative and stable brain signal patterns by imagery, which is challenging and is usually achieved over a very long training time (weeks/months). Here, we propose a human-machine joint learning framework to boost the learning process in endogenous BCIs, by guiding the user to generate brain signals towards an optimal distribution estimated by the decoder, given the historical brain signals of the user. To this end, we firstly model the human-machine joint learning process in a uniform formulation. Then a human-machine joint learning framework is proposed: 1) for the human side, we model the learning process in a sequential trial-and-error scenario and propose a novel ``copy/new'' feedback paradigm to help shape the signal generation of the subject toward the optimal distribution; 2) for the machine side, we propose a novel adaptive learning algorithm to learn an optimal signal distribution along with the subject's learning process. Specifically, the decoder reweighs the brain signals generated by the subject to focus more on ``good'' samples to cope with the learning process of the subject. Online and psuedo-online BCI experiments with 18 healthy subjects demonstrated the advantages of the proposed joint learning process over co-adaptive approaches in both learning efficiency and effectiveness.

cs.HC↗

arXiv2023

Deep Learning Applications Based on WISE Infrared Data: Classification of Stars, Galaxies and Quasars

The Wide-field Infrared Survey Explorer (WISE) has detected hundreds of millions of sources over the entire sky. However, classifying them reliably is a great challenge due to degeneracies in WISE multicolor space and low detection levels in its two longest-wavelength bandpasses. In this paper, the deep learning classification network, IICnet (Infrared Image Classification network), is designed to classify sources from WISE images to achieve a more accurate classification goal. IICnet shows good ability on the feature extraction of the WISE sources. Experiments demonstrates that the classification results of IICnet are superior to some other methods; it has obtained 96.2% accuracy for galaxies, 97.9% accuracy for quasars, and 96.4% accuracy for stars, and the Area Under Curve (AUC) of the IICnet classifier can reach more than 99%. In addition, the superiority of IICnet in processing infrared images has been demonstrated in the comparisons with VGG16, GoogleNet, ResNet34, MobileNet, EfficientNetV2, and RepVGG-fewer parameters and faster inference. The above proves that IICnet is an effective method to classify infrared sources.

astro-ph.IM↗

arXiv2023

Boosted ab initio Cryo-EM 3D Reconstruction with ACE-EM

The central problem in cryo-electron microscopy (cryo-EM) is to recover the 3D structure from noisy 2D projection images which requires estimating the missing projection angles (poses). Recent methods attempted to solve the 3D reconstruction problem with the autoencoder architecture, which suffers from the latent vector space sampling problem and frequently produces suboptimal pose inferences and inferior 3D reconstructions. Here we present an improved autoencoder architecture called ACE (Asymmetric Complementary autoEncoder), based on which we designed the ACE-EM method for cryo-EM 3D reconstructions. Compared to previous methods, ACE-EM reached higher pose space coverage within the same training time and boosted the reconstruction performance regardless of the choice of decoders. With this method, the Nyquist resolution (highest possible resolution) was reached for 3D reconstructions of both simulated and experimental cryo-EM datasets. Furthermore, ACE-EM is the only amortized inference method that reached the Nyquist resolution.

cs.CV↗

arXiv2023

3D Molecular Generation via Virtual Dynamics

Structure-based drug design, i.e., finding molecules with high affinities to the target protein pocket, is one of the most critical tasks in drug discovery. Traditional solutions, like virtual screening, require exhaustively searching on a large molecular database, which are inefficient and cannot return novel molecules beyond the database. The pocket-based 3D molecular generation model, i.e., directly generating a molecule with a 3D structure and binding position in the pocket, is a new promising way to address this issue. Herein, we propose VD-Gen, a novel pocket-based 3D molecular generation pipeline. VD-Gen consists of several carefully designed stages to generate fine-grained 3D molecules with binding positions in the pocket cavity end-to-end. Rather than directly generating or sampling atoms with 3D positions in the pocket like in early attempts, in VD-Gen, we first randomly initialize many virtual particles in the pocket; then iteratively move these virtual particles, making the distribution of virtual particles approximate the distribution of molecular atoms. After virtual particles are stabilized in 3D space, we extract a 3D molecule from them. Finally, we further refine atoms in the extracted molecule by iterative movement again, to get a high-quality 3D molecule, and predict a confidence score for it. Extensive experiment results on pocket-based molecular generation demonstrate that VD-Gen can generate novel 3D molecules to fill the target pocket cavity with high binding affinities, significantly outperforming previous baselines.

q-bio.BM↗

arXiv2023

Improving the precision of multiparameter estimation in the teleportation of qutrit under amplitude damping noise

Since the initial discovery of quantum teleportation, it is devoted to transferring unknown quantum states from one party to another distant partner. However, in the scenarios of remote sensing, what people truly care about is the information carried by certain parameters. The problem of multiparameter estimation in the framework of qutrit teleportation under amplitude damping (AD) noise is studied. Particularly, two schemes are proposed to battle against AD noise and enhance the precision of multiparameter estimation by utilizing weak measurement (WM) and environment-assisted measurement (EAM). For two-phase parameters encoded in a qutrit state, the analytical formulas of the quantum Fisher information matrix (QFIM) can be obtained. The results prove that the scheme of EAM outperforms the WM one in the improvements of both independent and simultaneous estimation precision. Remarkably, the EAM scheme can completely ensure the estimation precision against the contamination by AD noise. The reason should be attributed to the fact that EAM is carried out after the AD noise. Thus, it extracts information from both the system and the environment. The findings show that the techniques of WM and EAM are helpful for remote quantum sensing and can be generalized to other qutrit-based quantum information tasks under AD decoherence.

quant-ph↗

arXiv2022

Experimental Realization of the Rabi-Hubbard Model with Trapped Ions

Quantum simulation provides important tools in studying strongly correlated many-body systems with controllable parameters. As a hybrid of two fundamental models in quantum optics and in condensed matter physics, the Rabi-Hubbard model demonstrates rich physics through the competition between local spin-boson interactions and long-range boson hopping. Here we report an experimental realization of the Rabi-Hubbard model using up to $16$ trapped ions and present a controlled study of its equilibrium properties and quantum dynamics. We observe the ground-state quantum phase transition by slowly quenching the coupling strength, and measure the quantum dynamical evolution in various parameter regimes. With the magnetization and the spin-spin correlation as probes, we verify the prediction of the model Hamiltonian by comparing theoretical results in small system sizes with experimental observations. For larger-size systems of $16$ ions and $16$ phonon modes, the effective Hilbert space dimension exceeds $2^{57}$, whose dynamics is intractable for classical supercomputers.

quant-ph↗

arXiv2022

WaBERT: A Low-resource End-to-end Model for Spoken Language Understanding and Speech-to-BERT Alignment

Historically lower-level tasks such as automatic speech recognition (ASR) and speaker identification are the main focus in the speech field. Interest has been growing in higher-level spoken language understanding (SLU) tasks recently, like sentiment analysis (SA). However, improving performances on SLU tasks remains a big challenge. Basically, there are two main methods for SLU tasks: (1) Two-stage method, which uses a speech model to transfer speech to text, then uses a language model to get the results of downstream tasks; (2) One-stage method, which just fine-tunes a pre-trained speech model to fit in the downstream tasks. The first method loses emotional cues such as intonation, and causes recognition errors during ASR process, and the second one lacks necessary language knowledge. In this paper, we propose the Wave BERT (WaBERT), a novel end-to-end model combining the speech model and the language model for SLU tasks. WaBERT is based on the pre-trained speech and language model, hence training from scratch is not needed. We also set most parameters of WaBERT frozen during training. By introducing WaBERT, audio-specific information and language knowledge are integrated in the short-time and low-resource training process to improve results on the dev dataset of SLUE SA tasks by 1.15% of recall score and 0.82% of F1 score. Additionally, we modify the serial Continuous Integrate-and-Fire (CIF) mechanism to achieve the monotonic alignment between the speech and text modalities.

cs.CL↗

arXiv2021

A Vertically Resolved MSE Framework Highlights the Role of the Boundary Layer in Convective Self-Aggregation

Convective self-aggregation refers to a phenomenon in which random convection can self-organize into large-scale clusters over an ocean surface with uniform temperature in cloud-resolving models. Previous literature studies convective aggregation primarily by analyzing vertically integrated (VI) moist static energy (MSE) variance. That is the global MSE variance, including both the local MSE variance at a given altitude and the covariance of MSE anomalies between different altitudes. Here we present a vertically resolved (VR) MSE framework that focuses on the local MSE variance to study convective self-aggregation. Using a cloud-resolving simulation, we show that the development of self-aggregation is associated with an increase of local MSE variance, and that the diabatic and adiabatic generation of the MSE variance is mainly dominated by the boundary layer (BL). The results agree with recent numerical simulation results and the available potential energy analyses showing that the BL plays a key role in the development of self-aggregation. We further present a detailed comparison between the global and local MSE variance frameworks in their mathematical formulation and diagnostic results, highlighting their differences.

physics.ao-ph↗

arXiv2017

Optimization on fixed low latency implementation of GBT protocol in FPGA

In the upgrade of ATLAS experiment, the front-end electronics components are subjected to a large radiation background. Meanwhile high speed optical links are required for the data transmission between the on-detector and off-detector electronics. The GBT architecture and the Versatile Link (VL) project are designed by CERN to support the 4.8 Gbps line rate bidirectional high-speed data transmission which is called GBT link. In the ATLAS upgrade, besides the link with on-detector, the GBT link is also used between different off-detector systems. The GBTX ASIC is designed for the on-detector front-end, correspondingly for the off-detector electronics, the GBT architecture is implemented in Field Programmable Gate Arrays (FPGA). CERN launches the GBT-FPGA project to provide examples in different types of FPGA. In the ATLAS upgrade framework, the Front-End LInk eXchange (FELIX) system is used to interface the front-end electronics of several ATLAS subsystems. The GBT link is used between them, to transfer the detector data and the timing, trigger, control and monitoring information. The trigger signal distributed in the down-link from FELIX to the front-end requires a fixed and low latency. In this paper, several optimizations on the GBT-FPGA IP core are introduced, to achieve a lower fixed latency. For FELIX, a common firmware will be used to interface different front-ends with support of both GBT modes: the forward error correction mode and the wide mode. The modified GBT-FPGA core has the ability to switch between the GBT modes without FPGA reprogramming. The system clock distribution of the multi-channel FELIX firmware is also discussed in this paper.

physics.ins-det↗

arXiv2011

A Fault-Tolerant Emergency-Aware Access Control Scheme for Cyber-Physical Systems

Access control is an issue of paramount importance in cyber-physical systems (CPS). In this paper, an access control scheme, namely FEAC, is presented for CPS. FEAC can not only provide the ability to control access to data in normal situations, but also adaptively assign emergency-role and permissions to specific subjects and inform subjects without explicit access requests to handle emergency situations in a proactive manner. In FEAC, emergency-group and emergency-dependency are introduced. Emergencies are processed in sequence within the group and in parallel among groups. A priority and dependency model called PD-AGM is used to select optimal response-action execution path aiming to eliminate all emergencies that occurred within the system. Fault-tolerant access control polices are used to address failure in emergency management. A case study of the hospital medical care application shows the effectiveness of FEAC.

cs.NI↗

arXiv2011

A Hop-by-hop Cross-layer Congestion Control Scheme for Wireless Sensor Networks

Congestions in wireless sensor networks (WSNs) could potentially cause packet loss, throughput impairment and energy waste. To address this issue, a hop-by-hop cross-layer congestion control scheme (HCCC) built on contention-based MAC protocol is proposed in this paper. According to MAC-layer channel information including buffer occupancy ratio and congestion degree of local node, HCCC dynamically adjusts channel access priority in MAC layer and data transmission rate of the node to tackle the problem of congestion. Simulations have been conducted to compare HCCC against closely-related existing schemes. The results show that HCCC exhibits considerable superiority in terms of packets loss ratio, throughput and energy efficiency.

cs.NI↗

arXiv2011

Energy Efficient Ant Colony Algorithms for Data Aggregation in Wireless Sensor Networks

In this paper, a family of ant colony algorithms called DAACA for data aggregation has been presented which contains three phases: the initialization, packet transmission and operations on pheromones. After initialization, each node estimates the remaining energy and the amount of pheromones to compute the probabilities used for dynamically selecting the next hop. After certain rounds of transmissions, the pheromones adjustment is performed periodically, which combines the advantages of both global and local pheromones adjustment for evaporating or depositing pheromones. Four different pheromones adjustment strategies are designed to achieve the global optimal network lifetime, namely Basic-DAACA, ES-DAACA, MM-DAACA and ACS-DAACA. Compared with some other data aggregation algorithms, DAACA shows higher superiority on average degree of nodes, energy efficiency, prolonging the network lifetime, computation complexity and success ratio of one hop transmission. At last we analyze the characteristic of DAACA in the aspects of robustness, fault tolerance and scalability.

cs.NI↗

arXiv2011

A Group Key Management Protocol Based on Weight-Balanced 2-3 Tree for Wireless Sensor Networks

Multicast in Wireless Sensor Networks (WSNs) is an attractive mechanism for delivering data to multiple receivers as it saves bandwidth. To guarantee the security of multicast, the group key is used to encrypt and decrypt the packages. However, providing key management services in WSNs is complicated because sensor nodes possess limited resources of computing, storage and communication. To address the balance between security and limited resources, a multicast group key management protocol based on the weight-balanced 2-3 tree is proposed to generate, distribute, and update the group key securely and efficiently. The decentralized group key management method is employed. A weight-balanced 2-3 key tree is formed in every subgroup. Instead of using the conventional symmetric and non-symmetric encryption algorithms, the Maximum Distance Separable (MDS) code technique is used to distribute the multicast key dynamically. During the key updating, a series of adjustment rules are summarized to keep the tree weight-balanced, where pseudo-nodes as leaves are added to reduce the computation and communication complexity. Compared with some other group key management protocols, our scheme shows higher superiority on security and performance.

cs.CR↗

arXiv2010

A Clustering-based Location Privacy Protection Scheme for Pervasive Computing

In pervasive computing environments, Location- Based Services (LBSs) are becoming increasingly important due to continuous advances in mobile networks and positioning technologies. Nevertheless, the wide deployment of LBSs can jeopardize the location privacy of mobile users. Consequently, providing safeguards for location privacy of mobile users against being attacked is an important research issue. In this paper a new scheme for safeguarding location privacy is proposed. Our approach supports location K-anonymity for a wide range of mobile users with their own desired anonymity levels by clustering. The whole area of all users is divided into clusters recursively in order to get the Minimum Bounding Rectangle (MBR). The exact location information of a user is replaced by his MBR. Privacy analysis shows that our approach can achieve high resilience to location privacy threats and provide more privacy than users expect. Complexity analysis shows clusters can be adjusted in real time as mobile users join or leave. Moreover, the clustering algorithms possess strong robustness.

cs.CR↗

arXiv2010

iCare: A Mobile Health Monitoring System for the Elderly

This paper describes a mobile health monitoring system called iCare for the elderly. We use wireless body sensors and smart phones to monitor the wellbeing of the elderly. It can offer remote monitoring for the elderly anytime anywhere and provide tailored services for each person based on their personal health condition. When detecting an emergency, the smart phone will automatically alert pre-assigned people who could be the old people's family and friends, and call the ambulance of the emergency centre. It also acts as the personal health information system and the medical guidance which offers one communication platform and the medical knowledge database so that the family and friends of the served people can cooperate with doctors to take care of him/her. The system also features some unique functions that cater to the living demands of the elderly, including regular reminder, quick alarm, medical guidance, etc. iCare is not only a real-time health monitoring system for the elderly, but also a living assistant which can make their lives more convenient and comfortable.

cs.OH↗