arXiv subjects

Jing Qi

Publications and source records attributed to Jing Qi.

9 recordsLinked to original sources

arXiv2026

Multi-Faceted Continual Knowledge Graph Embedding for Semantic-Aware Link Prediction

Continual Knowledge Graph Embedding (CKGE) aims to continually learn embeddings for new knowledge, i.e., entities and relations, while retaining previously acquired knowledge. Most existing CKGE methods mitigate catastrophic forgetting via regularization or replaying old knowledge. They conflate new and old knowledge of an entity within the same embedding space to seek a balance between them. However, entities inherently exhibit multi-faceted semantics that evolve dynamically as their relational contexts change over time. A shared embedding fails to capture and distinguish these temporal semantic variations, degrading lifelong link prediction accuracy across snapshots. To address this, we propose a Multi-Faceted CKGE framework (MF-CKGE) for semantic-aware link prediction. During offline learning, MF-CKGE separates temporal old and new knowledge into distinct embedding spaces to prevent knowledge entanglement and employs semantic decoupling to reduce semantic redundancy, thereby improving space efficiency. During online inference, MF-CKGE adaptively identifies semantically query-relevant entity embeddings by quantifying their semantic importance, reducing interference from query-irrelevant noise. Experiments on eight datasets show that MF-CKGE achieves an average (maximum) improvement of 1.7% (2.7%) and 1.4% (3.8%) in MRR and Hits@10, respectively, over the best baseline. Our source code and datasets are available at: https://anonymous.4open.science/r/MF-CKGE-04E5.

cs.IR↗

arXiv2024

A Joint Representation Using Continuous and Discrete Features for Cardiovascular Diseases Risk Prediction on Chest CT Scans

Cardiovascular diseases (CVD) remain a leading health concern and contribute significantly to global mortality rates. While clinical advancements have led to a decline in CVD mortality, accurately identifying individuals who could benefit from preventive interventions remains an unsolved challenge in preventive cardiology. Current CVD risk prediction models, recommended by guidelines, are based on limited traditional risk factors or use CT imaging to acquire quantitative biomarkers, and still have limitations in predictive accuracy and applicability. On the other hand, end-to-end trained CVD risk prediction methods leveraging deep learning on CT images often fail to provide transparent and explainable decision grounds for assisting physicians. In this work, we proposed a novel joint representation that integrates discrete quantitative biomarkers and continuous deep features extracted from chest CT scans. Our approach initiated with a deep CVD risk classification model by capturing comprehensive continuous deep learning features while jointly obtaining currently clinical-established quantitative biomarkers via segmentation models. In the feature joint representation stage, we use an instance-wise feature-gated mechanism to align the continuous and discrete features, followed by a soft instance-wise feature interaction mechanism fostering independent and effective feature interaction for the final CVD risk prediction. Our method substantially improves CVD risk predictive performance and offers individual contribution analysis of each biomarker, which is important in assisting physicians' decision-making processes. We validated our method on a public chest low-dose CT dataset and a private external chest standard-dose CT patient cohort of 17,207 CT volumes from 6,393 unique subjects, and demonstrated superior predictive performance, achieving AUCs of 0.875 and 0.843, respectively.

eess.IV↗

arXiv2024

SGDM: Static-Guided Dynamic Module Make Stronger Visual Models

The spatial attention mechanism has been widely used to improve object detection performance. However, its operation is currently limited to static convolutions lacking content-adaptive features. This paper innovatively approaches from the perspective of dynamic convolution. We propose Razor Dynamic Convolution (RDConv) to address thetwo flaws in dynamic weight convolution, making it hard to implement in spatial mechanism: 1) it is computation-heavy; 2) when generating weights, spatial information is disregarded. Firstly, by using Razor Operation to generate certain features, we vastly reduce the parameters of the entire dynamic convolution operation. Secondly, we added a spatial branch inside RDConv to generate convolutional kernel parameters with richer spatial information. Embedding dynamic convolution will also bring the problem of sensitivity to high-frequency noise. We propose the Static-Guided Dynamic Module (SGDM) to address this limitation. By using SGDM, we utilize a set of asymmetric static convolution kernel parameters to guide the construction of dynamic convolution. We introduce the mechanism of shared weights in static convolution to solve the problem of dynamic convolution being sensitive to high-frequency noise. Extensive experiments illustrate that multiple different object detection backbones equipped with SGDM achieve a highly competitive boost in performance(e.g., +4% mAP with YOLOv5n on VOC and +1.7% mAP with YOLOv8n on COCO) with negligible parameter increase(i.e., +0.33M on YOLOv5n and +0.19M on YOLOv8n).

cs.CV↗

arXiv2023

Evidence for a conical spin spiral state in the Mn triple-layer on W(001): spin-polarized scanning tunneling microscopy and first-principles calculations

The spin structure of a Mn triple layer grown pseudomorphically on surfaces is studied using spin-polarized scanning tunneling microscopy (SP-STM) and density functional theory (DFT). In SP-STM images a c$(4 \times 2)$ super structure is found. The magnetic origin of this contrast is verified by contrast reversal and using the c$(2 \times 2)$ AFM state of the Mn double layer as a reference. SP-STM simulations show that this contrast can be explained by a spin spiral propagating along the [110] direction with an angle close to $90^\circ$ between magnetic moments of adjacent Mn rows. To understand the origin of this spin structure, DFT calculations have been performed for a large number of competing collinear and non-collinear magnetic states including the effect of spin-orbit oupling (SOC). Surprisingly, a collinear state in which the magnetic moments of top and central Mn layer are aligned antiparallel and those of the bottom Mn layer are aligned parallel to the central layer is the energetically lowest state. We show that in this so-called "up-down-down" ($\uparrow \downarrow \downarrow$) state the magnetic moments in the Mn bottom layer are only induced by those of the central Mn layer. Flat spin spirals propagating either in one, two, or all Mn layers are shown to be energetically unfavorable to the collinear $\uparrow \downarrow \downarrow$ state even upon including the Dzyaloshinskii-Moriya interaction (DMI). However, conical spin spirals with a small opening angle of about $10^\circ$ are only slightly energetically unfavorable within DFT and could explain the experimental observations. Surprisingly, the DFT energy dispersion of conical spin spirals including SOC cannot be explained if only the DMI is taken into account. Therefore, higher-order interactions such as chiral biquadratic terms need to be considered which could explain the stabilization of a conical spin spiral state.

cond-mat.mtrl-sci↗

arXiv2022

Reversible Tuning of Collinear versus Chiral Magnetic Order by Chemical Stimulus

The Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction mediates collinear magnetic interactions via the conduction electrons of a non-magnetic spacer, resulting in a ferro- or antiferromagnetic magnetization in magnetic multilayers. The resulting spin-polarized charge transport effects have found numerous applications. Recently it has been discovered that heavy non-magnetic spacers are able to mediate an indirect magnetic coupling that is non-collinear and chiral. This Dzyaloshinskii-Moriya-enhanced RKKY (DME-RKKY) interaction causes the emergence of a variety of interesting magnetic structures, such as skyrmions and spin spirals. Applications using these magnetic quasi-particles require a thorough understanding and fine-tuning of the balance between the Dzyaloshinskii-Moriya interaction and other magnetic interactions, e.g., the exchange interaction and magnetic anisotropy contributions. Here, we show by spin-polarized scanning tunneling microscopy that the spin structure of manganese oxide chains on Ir(001) can reproducibly be switched from chiral to collinear antiferromagnetic interchain interactions by increasing the oxidation state of MnO$_2$ while the reverse process can be induced by thermal reduction. The underlying structural change is revealed by low-energy electron diffraction intensity data (LEED-IV) analysis. Density functional theory calculations suggest that the magnetic transition may be caused by a significant increase of the Heisenberg exchange upon oxidation.

cond-mat.mtrl-sci↗

arXiv2021

Extrinsic ferroelectricity originated from oxygen vacancy drift in HfO2-based films

It is generally accepted that oxygen vacancies play a central role in the emergence of ferroelectricity for HfO2-based materials, but the underlying mechanism still remains elusive. Herein, starting from the basic characterization circuit, we propose that the observed ferroelectricity is extrinsic. A key finding is that charged oxygen vacancies oscillate within the sample under repeated electric pulses, yielding a nonlinear current which behaves similarly to the polarization current for a normal ferroelectric. This unwanted current signal results in a ferroelectric-like hysteresis loop with both remnant polarization and coercive field in good agreements with experimental values, given a charged oxygen vacancy concentration in the vicinity of 1*10^20/cm^3. Moreover, it is possible to exploit this mechanism to reproduce the effects of wake-up, split-up and limited endurance that are of crucial relevance for the device applications.

cond-mat.mtrl-sci↗

arXiv2020

Adversarial Data Encryption

In the big data era, many organizations face the dilemma of data sharing. Regular data sharing is often necessary for human-centered discussion and communication, especially in medical scenarios. However, unprotected data sharing may also lead to data leakage. Inspired by adversarial attack, we propose a method for data encryption, so that for human beings the encrypted data look identical to the original version, but for machine learning methods they are misleading. To show the effectiveness of our method, we collaborate with the Beijing Tiantan Hospital, which has a world leading neurological center. We invite $3$ doctors to manually inspect our encryption method based on real world medical images. The results show that the encrypted images can be used for diagnosis by the doctors, but not by machine learning methods.

cs.LG↗

arXiv2015

Characterization of a circular optical nanoantenna by nonlinear photoemission electron microscopy

We report on the investigation of an advanced circular plasmonic nanoantenna under ultrafast excitation using nonlinear photoemission electron microscopy (PEEM) under near-normal incidence. The circular nanoantenna is enhanced in its performance by a supporting grating and milled out from a gold film. The considered antenna shows a sophisticated physical resonance behavior that is ideal to demonstrate the possibilities of PEEM for the experimental investigations of plasmonic effects on the nanoscale. Field profiles of the antenna resonance for both possible linear polarizations of the incident field are measured with high spatial resolution. In addition, outward propagating Hankel plasmons, which are also excited by the structure, are measured and analyzed. We compare our findings to measurements of an isolated plasmonic nanodisc resonator and scanning near-field optical microscopy (SNOM) measurements of both structures. All results are in very good agreement with numerical simulations as well as analytial models that are also discussed in our paper.

physics.optics↗

arXiv2012

Circular Optical Nanoantennas: An Analytical Theory

An entirely analytical theory is provided for describing the resonance properties of optical nanoantennas made of a stack of homogeneous discs, i.e. circular patch nanoantennas. It consists in analytically calculating the phase accumulation of surface plasmon polaritons across the resonator and an additional contribution from the complex reflection coefficient at the antenna termination. This makes the theory self-contained with no need for fitting parameters. The very antenna resonances are then explained by a simple Fabry-Perot resonator model. Predictions are compared to rigorous simulations and show excellent agreement. Using this analytical model, circular antennas can be tuned by varying the composition of the stack.

physics.optics↗