arXiv subjects

Lin Li

Publications and source records attributed to Lin Li.

At least 271 records · Page 15Linked to original sources

arXiv2023

A Unified BEV Model for Joint Learning of 3D Local Features and Overlap Estimation

Pairwise point cloud registration is a critical task for many applications, which heavily depends on finding correct correspondences from the two point clouds. However, the low overlap between input point clouds causes the registration to fail easily, leading to mistaken overlapping and mismatched correspondences, especially in scenes where non-overlapping regions contain similar structures. In this paper, we present a unified bird's-eye view (BEV) model for jointly learning of 3D local features and overlap estimation to fulfill pairwise registration and loop closure. Feature description is performed by a sparse UNet-like network based on BEV representation, and 3D keypoints are extracted by a detection head for 2D locations, and a regression head for heights. For overlap detection, a cross-attention module is applied for interacting contextual information of input point clouds, followed by a classification head to estimate the overlapping region. We evaluate our unified model extensively on the KITTI dataset and Apollo-SouthBay dataset. The experiments demonstrate that our method significantly outperforms existing methods on overlap estimation, especially in scenes with small overlaps. It also achieves top registration performance on both datasets in terms of translation and rotation errors.

cs.CV↗

arXiv2023

Instance-wise or Class-wise? A Tale of Neighbor Shapley for Concept-based Explanation

Deep neural networks have demonstrated remarkable performance in many data-driven and prediction-oriented applications, and sometimes even perform better than humans. However, their most significant drawback is the lack of interpretability, which makes them less attractive in many real-world applications. When relating to the moral problem or the environmental factors that are uncertain such as crime judgment, financial analysis, and medical diagnosis, it is essential to mine the evidence for the model's prediction (interpret model knowledge) to convince humans. Thus, investigating how to interpret model knowledge is of paramount importance for both academic research and real applications.

cs.LG↗

arXiv2023

An Efficient Spectral Trust-Region Deflation Method for Multiple Solutions

It is quite common that a nonlinear partial differential equation (PDE) admits multiple distinct solutions and each solution may carry a unique physical meaning. One typical approach for finding multiple solutions is to use the Newton method with different initial guesses that ideally fall into the basins of attraction confining the solutions. In this paper, we propose a fast and accurate numerical method for multiple solutions comprised of three ingredients: (i) a well-designed spectral-Galerkin discretization of the underlying PDE leading to a nonlinear algebraic system (NLAS) with multiple solutions; (ii) an effective deflation technique to eliminate a known (founded) solution from the other unknown solutions leading to deflated NLAS; and (iii) a viable nonlinear least-squares and trust-region (LSTR) method for solving the NLAS and the deflated NLAS to find the multiple solutions sequentially one by one. We demonstrate through ample examples of differential equations and comparison with relevant existing approaches that the spectral LSTR-Deflation method has the merits: (i) it is quite flexible in choosing initial values, even starting from the same initial guess for finding all multiple solutions; (ii) it guarantees high-order accuracy; and (iii) it is quite fast to locate multiple distinct solutions and explore new solutions which are not reported in literature.

math.NA↗

arXiv2023

Tunable correlation effects of magnetic impurities by the cubic Rashba spin-orbit couplings

We theoretically study the influence of the $k$-cubic Rashba spin-orbit coupling (SOC) on the correlation effects of magnetic impurities by combining the variational method and the Hirsch-Fye quantum Monte Carlo (HFQMC) simulations. Markedly different from the normal $k$-linear Rashba SOC, even a small cubic Rashba term can greatly alter the band structure and induce a Van Hove singularity in a wide range of energy, thus the single impurity local moment becomes largely tunable. The cubic Rashba SOC adopted in this work breaks the rotational symmetry, but the host material is still invariant under the operations $\mathcal{R}^z(π)$, $\mathcal{IR}^z(π/2)$, $\mathcal{M}_{xz}$, $\mathcal{M}_{yz}$, where $\mathcal{R}^z(θ)$ is the rotation of angle $θ$ about the $z$-axis, $\mathcal{I}$ is the inversion operator and $\mathcal{M}_{xz}$ ($\mathcal{M}_{yz}$) is the mirror reflection about the $x$-$z$ ($y$-$z$) principal plane. Saliently, various components of spin-spin correlation between the single magnetic impurity and the conduction electrons show three- or six-fold rotational symmetry. This unique feature is due to the triple winding of the spins with a $2π$ rotation of $\mathbf{k}$, which is a hallmark of the cubic Rashba effect and can possibly be an identifier to distinguish the cubic Rashba SOC from the normal $k$-linear Rashba term in experiments. Although the cubic Rashba term drastically alters the electronic properties of the host, we find that the spatial decay rate of the spin-spin correlation function remains essentially unchanged. Moreover, the carrier-mediated Ruderman-Kittel-Kasuya-Yosida interactions between two magnetic impurities show twisted features, the ferromagnetic diagonal terms dominate when two magnetic impurities are very close, but the off-diagonal terms become important at long distances.

cond-mat.str-el↗

arXiv2023

Data Augmentation Alone Can Improve Adversarial Training

Adversarial training suffers from the issue of robust overfitting, which seriously impairs its generalization performance. Data augmentation, which is effective at preventing overfitting in standard training, has been observed by many previous works to be ineffective in mitigating overfitting in adversarial training. This work proves that, contrary to previous findings, data augmentation alone can significantly boost accuracy and robustness in adversarial training. We find that the hardness and the diversity of data augmentation are important factors in combating robust overfitting. In general, diversity can improve both accuracy and robustness, while hardness can boost robustness at the cost of accuracy within a certain limit and degrade them both over that limit. To mitigate robust overfitting, we first propose a new crop transformation, Cropshift, which has improved diversity compared to the conventional one (Padcrop). We then propose a new data augmentation scheme, based on Cropshift, with much improved diversity and well-balanced hardness. Empirically, our augmentation method achieves the state-of-the-art accuracy and robustness for data augmentations in adversarial training. Furthermore, when combined with weight averaging it matches, or even exceeds, the performance of the best contemporary regularization methods for alleviating robust overfitting. Code is available at: https://github.com/TreeLLi/DA-Alone-Improves-AT.

cs.CV↗

arXiv2023

Remote patient monitoring using artificial intelligence: Current state, applications, and challenges

The adoption of artificial intelligence (AI) in healthcare is growing rapidly. Remote patient monitoring (RPM) is one of the common healthcare applications that assist doctors to monitor patients with chronic or acute illness at remote locations, elderly people in-home care, and even hospitalized patients. The reliability of manual patient monitoring systems depends on staff time management which is dependent on their workload. Conventional patient monitoring involves invasive approaches which require skin contact to monitor health status. This study aims to do a comprehensive review of RPM systems including adopted advanced technologies, AI impact on RPM, challenges and trends in AI-enabled RPM. This review explores the benefits and challenges of patient-centric RPM architectures enabled with Internet of Things wearable devices and sensors using the cloud, fog, edge, and blockchain technologies. The role of AI in RPM ranges from physical activity classification to chronic disease monitoring and vital signs monitoring in emergency settings. This review results show that AI-enabled RPM architectures have transformed healthcare monitoring applications because of their ability to detect early deterioration in patients' health, personalize individual patient health parameter monitoring using federated learning, and learn human behavior patterns using techniques such as reinforcement learning. This review discusses the challenges and trends to adopt AI to RPM systems and implementation issues. The future directions of AI in RPM applications are analyzed based on the challenges and trends

cs.CY↗

arXiv2023

Towards A Unified Conformer Structure: from ASR to ASV Task

Transformer has achieved extraordinary performance in Natural Language Processing and Computer Vision tasks thanks to its powerful self-attention mechanism, and its variant Conformer has become a state-of-the-art architecture in the field of Automatic Speech Recognition (ASR). However, the main-stream architecture for Automatic Speaker Verification (ASV) is convolutional Neural Networks, and there is still much room for research on the Conformer based ASV. In this paper, firstly, we modify the Conformer architecture from ASR to ASV with very minor changes. Length-Scaled Attention (LSA) method and Sharpness-Aware Minimizationis (SAM) are adopted to improve model generalization. Experiments conducted on VoxCeleb and CN-Celeb show that our Conformer based ASV achieves competitive performance compared with the popular ECAPA-TDNN. Secondly, inspired by the transfer learning strategy, ASV Conformer is natural to be initialized from the pretrained ASR model. Via parameter transferring, self-attention mechanism could better focus on the relationship between sequence features, brings about 11% relative improvement in EER on test set of VoxCeleb and CN-Celeb, which reveals the potential of Conformer to unify ASV and ASR task. Finally, we provide a runtime in ASV-Subtools to evaluate its inference speed in production scenario. Our code is released at https://github.com/Snowdar/asv-subtools/tree/master/doc/papers/conformer.md.

eess.AS↗

arXiv2022

A scheme for excitation of thorium-229 nuclei based on the electronic bridge excitaion

Thorium-229 possesses the lowest nuclear first excited state with an energy of about 8 eV. The extremely narrow linewidth of the nuclear first excited state with the uncertainty of 53 THz prevents the direct laser excitation and the realization of the nuclear clock. We present a proposal using the Coulomb crystal of a linear chain formed by the $^{229}$Th$^{3+}$ ions, the nuclei of $^{229}$Th$^{3+}$ ions in the ion trap are excited by the electronic bridge (EB) process. The 7$P_{1/2}$ state of the thorium-229 nuclear ground state is chosen for the EB excitation. Using the two-level optical Bloch equation under experimental conditions, we calculate that 2 out of 36 prepared thorium ions in the Coulomb crystal can be excited to the nuclear first excited state, and it takes about 2 hours to scan over the uncertainty of 0.22 eV. Taking the advantage of transition enhancement of the EB and the long stability of the Coulomb crystal, the energy uncertainty of the first excited state can be limited to the order of 1 GHz.

physics.atom-ph↗

arXiv2022

Understanding and Combating Robust Overfitting via Input Loss Landscape Analysis and Regularization

Adversarial training is widely used to improve the robustness of deep neural networks to adversarial attack. However, adversarial training is prone to overfitting, and the cause is far from clear. This work sheds light on the mechanisms underlying overfitting through analyzing the loss landscape w.r.t. the input. We find that robust overfitting results from standard training, specifically the minimization of the clean loss, and can be mitigated by regularization of the loss gradients. Moreover, we find that robust overfitting turns severer during adversarial training partially because the gradient regularization effect of adversarial training becomes weaker due to the increase in the loss landscapes curvature. To improve robust generalization, we propose a new regularizer to smooth the loss landscape by penalizing the weighted logits variation along the adversarial direction. Our method significantly mitigates robust overfitting and achieves the highest robustness and efficiency compared to similar previous methods. Code is available at https://github.com/TreeLLi/Combating-RO-AdvLC.

cs.LG↗

arXiv2022

Proximity effect in PbTe-Pb hybrid nanowire Josephson junctions

Semiconductor-superconductor hybrid nanowires are a leading material platform for the realization of Majorana zero modes. The semiconductors in previous studies are dominantly InAs or InSb. In this work, we show the induced superconductivity in a PbTe nanowire epitaxially coupled to a superconductor Pb. The Josephson junction device based on this hybrid reveals a gate-tunable supercurrent in the open regime and a hard superconducting gap in the tunneling regime. By demonstrating the superconducting proximity effect, our result can enable Majorana searches and other applications like gate-tunable qubits in a new semiconductor system.

cond-mat.mes-hall↗

arXiv2022

Graph Contrastive Learning for Materials

Recent work has shown the potential of graph neural networks to efficiently predict material properties, enabling high-throughput screening of materials. Training these models, however, often requires large quantities of labelled data, obtained via costly methods such as ab initio calculations or experimental evaluation. By leveraging a series of material-specific transformations, we introduce CrystalCLR, a framework for constrastive learning of representations with crystal graph neural networks. With the addition of a novel loss function, our framework is able to learn representations competitive with engineered fingerprinting methods. We also demonstrate that via model finetuning, contrastive pretraining can improve the performance of graph neural networks for prediction of material properties and significantly outperform traditional ML models that use engineered fingerprints. Lastly, we observe that CrystalCLR produces material representations that form clusters by compound class.

cs.LG↗

arXiv2022

A multi-band atomic candle with microwave-dressed Rydberg atoms

Stabilizing important physical quantities to atom-based standards lies at the heart of modern atomic, molecular and optical physics, and is widely applied to the field of precision metrology. Of particular importance is the atom-based microwave field amplitude stabilizer, the so-called atomic candle. Previous atomic candles are realized with atoms in their ground state, and hence suffer from the lack of frequency band tunability and small stabilization bandwidth, severely limiting their development and potential applications. To tackle these limitations, we employ microwave-dressed Rydberg atoms to realize a novel atomic candle that features multi-band frequency tunability and large stabilization bandwidth. We demonstrate amplitude stabilization of microwave field from C-band to Ka-band, which could be extended to quasi-DC and terahertz fields by exploring abundant Rydberg levels. Our atomic candle achieves stabilization bandwidth of 100 Hz, outperforming previous ones by more than two orders of magnitude. Our simulation indicates the stabilization bandwidth can be further increased up to 100 kHz. Our work paves a route to develop novel electric field control and applications with a noise-resilient, miniaturized, sensitive and broadband atomic candle.

physics.atom-ph↗

arXiv2022

Antibody Representation Learning for Drug Discovery

Therapeutic antibody development has become an increasingly popular approach for drug development. To date, antibody therapeutics are largely developed using large scale experimental screens of antibody libraries containing hundreds of millions of antibody sequences. The high cost and difficulty of developing therapeutic antibodies create a pressing need for computational methods to predict antibody properties and create bespoke designs. However, the relationship between antibody sequence and activity is a complex physical process and traditional iterative design approaches rely on large scale assays and random mutagenesis. Deep learning methods have emerged as a promising way to learn antibody property predictors, but predicting antibody properties and target-specific activities depends critically on the choice of antibody representations and data linking sequences to properties is often limited. Existing works have not yet investigated the value, limitations and opportunities of these methods in application to antibody-based drug discovery. In this paper, we present results on a novel SARS-CoV-2 antibody binding dataset and an additional benchmark dataset. We compare three classes of models: conventional statistical sequence models, supervised learning on each dataset independently, and fine-tuning an antibody specific pre-trained language model. Experimental results suggest that self-supervised pretraining of feature representation consistently offers significant improvement in over previous approaches. We also investigate the impact of data size on the model performance, and discuss challenges and opportunities that the machine learning community can address to advance in silico engineering and design of therapeutic antibodies.

q-bio.QM↗

arXiv2022

Spatial-aware Speaker Diarization for Multi-channel Multi-party Meeting

This paper describes a spatial-aware speaker diarization system for the multi-channel multi-party meeting. The diarization system obtains direction information of speaker by microphone array. Speaker spatial embedding is generated by xvector and s-vector derived from superdirective beamforming (SDB) which makes the embedding more robust. Specifically, we propose a novel multi-channel sequence-to-sequence neural network architecture named discriminative multi-stream neural network (DMSNet) which consists of attention superdirective beamforming (ASDB) block and Conformer encoder. The proposed ASDB is a self-adapted channel-wise block that extracts the latent spatial features of array audios by modeling interdependencies between channels. We explore DMSNet to address overlapped speech problem on multi-channel audio and achieve 93.53% accuracy on evaluation set. By performing DMSNet based overlapped speech detection (OSD) module, the diarization error rate (DER) of cluster-based diarization system decrease significantly from 13.45% to 7.64%.

eess.AS↗

arXiv2022

Entanglement filter with Rydberg atoms

Devices capable of deterministically manipulating the photonic entanglement are of paramount importance, since photons are the ideal messengers for quantum information. Here, we report a Rydberg-atom-based entanglement filter that preserves the desired photonic entangled state and deterministically blocks the transmission of the unwanted ones. Photonic entanglement with near-unity fidelity can be extracted from an input state with an arbitrarily low initial fidelity. The protocol is inherently robust, and succeeds both in the Rydberg blockade regime and in the interaction-induced dissipation regime. Such an entanglement filter opens new routes toward scalable photonic quantum information processing with multiple ensembles of Rydberg atoms.

physics.atom-ph↗

arXiv2022

Label Semantic Knowledge Distillation for Unbiased Scene Graph Generation

The Scene Graph Generation (SGG) task aims to detect all the objects and their pairwise visual relationships in a given image. Although SGG has achieved remarkable progress over the last few years, almost all existing SGG models follow the same training paradigm: they treat both object and predicate classification in SGG as a single-label classification problem, and the ground-truths are one-hot target labels. However, this prevalent training paradigm has overlooked two characteristics of current SGG datasets: 1) For positive samples, some specific subject-object instances may have multiple reasonable predicates. 2) For negative samples, there are numerous missing annotations. Regardless of the two characteristics, SGG models are easy to be confused and make wrong predictions. To this end, we propose a novel model-agnostic Label Semantic Knowledge Distillation (LS-KD) for unbiased SGG. Specifically, LS-KD dynamically generates a soft label for each subject-object instance by fusing a predicted Label Semantic Distribution (LSD) with its original one-hot target label. LSD reflects the correlations between this instance and multiple predicate categories. Meanwhile, we propose two different strategies to predict LSD: iterative self-KD and synchronous self-KD. Extensive ablations and results on three SGG tasks have attested to the superiority and generality of our proposed LS-KD, which can consistently achieve decent trade-off performance between different predicate categories.

cs.CV↗

arXiv2022

PanGu-Coder: Program Synthesis with Function-Level Language Modeling

We present PanGu-Coder, a pretrained decoder-only language model adopting the PanGu-Alpha architecture for text-to-code generation, i.e. the synthesis of programming language solutions given a natural language problem description. We train PanGu-Coder using a two-stage strategy: the first stage employs Causal Language Modelling (CLM) to pre-train on raw programming language data, while the second stage uses a combination of Causal Language Modelling and Masked Language Modelling (MLM) training objectives that focus on the downstream task of text-to-code generation and train on loosely curated pairs of natural language program definitions and code functions. Finally, we discuss PanGu-Coder-FT, which is fine-tuned on a combination of competitive programming problems and code with continuous integration tests. We evaluate PanGu-Coder with a focus on whether it generates functionally correct programs and demonstrate that it achieves equivalent or better performance than similarly sized models, such as CodeX, while attending a smaller context window and training on less data.

cs.LG↗

arXiv2022

Short-range order and its impacts on the BCC NbMoTaW multi-principal element alloy by the machine-learning potential

We employ a machine-learning force field, trained by a neural network (NN) with bispectrum coefficients as descriptors, to investigate the short-range order (SRO) influences on the BCC NbMoTaW alloy strengthening mechanism. The NN interatomic potential provides a transferable force field with density functional theory accuracy. This novel NN potential is applied to elucidate the SRO effects on the elasticity, vibrational modes, plasticity, and strength in the NbMoTaW multi-principal element alloy (MPEA). The results show the strong attraction among Mo-Ta pairs forming the local ordered B2 structures, which could be tuned via temperature and improved by Nb content. SRO increases the elastic constants and high-frequency phonon modes as well as introduces extra lattice friction of dislocation motion. This approach enables a rapid compositional screening, paves the way for computation-guided materials design of new MPEAs with better performance, and opens avenues for tuning the mechanical properties by processing optimization.

cond-mat.mtrl-sci↗