arXiv subjects

Rui Cai

Publications and source records attributed to Rui Cai.

28 records · Page 2Linked to original sources

arXiv2024

Zero-Shot Relational Learning for Multimodal Knowledge Graphs

Relational learning is an essential task in the domain of knowledge representation, particularly in knowledge graph completion (KGC). While relational learning in traditional single-modal settings has been extensively studied, exploring it within a multimodal KGC context presents distinct challenges and opportunities. One of the major challenges is inference on newly discovered relations without any associated training data. This zero-shot relational learning scenario poses unique requirements for multimodal KGC, i.e., utilizing multimodality to facilitate relational learning.However, existing works fail to support the leverage of multimodal information and leave the problem unexplored. In this paper, we propose a novel end-to-end framework, consisting of three components, i.e., multimodal learner, structure consolidator, and relation embedding generator, to integrate diverse multimodal information and knowledge graph structures to facilitate the zero-shot relational learning. Evaluation results on three multimodal knowledge graphs demonstrate the superior performance of our proposed method.

cs.LG↗

arXiv2023

Masked Face Dataset Generation and Masked Face Recognition

In the post-pandemic era, wearing face masks has posed great challenge to the ordinary face recognition. In the previous study, researchers has applied pretrained VGG16, and ResNet50 to extract features on the elaborate curated existing masked face recognition (MFR) datasets, RMFRD and SMFRD. To make the model more adaptable to the real world situation where the sample size is smaller and the camera environment has greater changes, we created a more challenging masked face dataset ourselves, by selecting 50 identities with 1702 images from Labelled Faces in the Wild (LFW) Dataset, and simulated face masks through key point detection. The another part of our study is to solve the masked face recognition problem, and we chose models by referring to the former state of the art results, instead of directly using pretrained models, we fine tuned the model on our new dataset and use the last linear layer to do the classification directly. Furthermore, we proposed using data augmentation strategy to further increase the test accuracy, and fine tuned a new networks beyond the former study, one of the most SOTA networks, Inception ResNet v1. The best test accuracy on 50 identity MFR has achieved 95%.

cs.CV↗

arXiv2023

Transformation of hexagonal Lu to cubic LuH$_{2+x}$ single-crystalline films

With the recent report of near ambient superconductivity at room temperature in the N-doped lutetium hydride (Lu-H-N) system, the understanding of cubic Lu-H compounds has attracted worldwide attention. Generally, compared to polycrystal structures with non-negligible impurities, the single-crystalline form of materials with high purity can provide an opportunity to show their hidden properties. However, the experimental synthesis of single-crystalline cubic Lu-H compounds has not been reported thus far. Here, we developed an easy way to synthesize highly pure LuH$_{2+x}$ single-crystalline films by the post-annealing of Lu single-crystalline films (purity of 99.99%) in the H$_2$ atmosphere. The crystal and electronic structures of films were characterized by X-ray diffraction, Raman spectroscopy, and electrical transport. Interestingly, Lu films are silver-white and metallic, whereas their transformed LuH$_{2+x}$ films become purple-red and insulating, indicating the formation of an unreported electronic state of Lu-H compounds possibly. Our work provides a novel route to synthesize and explore more single-crystalline Lu-H compounds.

cond-mat.mtrl-sci↗

arXiv2022

Cross-Lingual Cross-Modal Retrieval with Noise-Robust Learning

Despite the recent developments in the field of cross-modal retrieval, there has been less research focusing on low-resource languages due to the lack of manually annotated datasets. In this paper, we propose a noise-robust cross-lingual cross-modal retrieval method for low-resource languages. To this end, we use Machine Translation (MT) to construct pseudo-parallel sentence pairs for low-resource languages. However, as MT is not perfect, it tends to introduce noise during translation, rendering textual embeddings corrupted and thereby compromising the retrieval performance. To alleviate this, we introduce a multi-view self-distillation method to learn noise-robust target-language representations, which employs a cross-attention module to generate soft pseudo-targets to provide direct supervision from the similarity-based view and feature-based view. Besides, inspired by the back-translation in unsupervised MT, we minimize the semantic discrepancies between origin sentences and back-translated sentences to further improve the noise robustness of the textual encoder. Extensive experiments are conducted on three video-text and image-text cross-modal retrieval benchmarks across different languages, and the results demonstrate that our method significantly improves the overall performance without using extra human-labeled data. In addition, equipped with a pre-trained visual encoder from a recent vision-and-language pre-training framework, i.e., CLIP, our model achieves a significant performance gain, showing that our method is compatible with popular pre-training models. Code and data are available at https://github.com/HuiGuanLab/nrccr.

cs.CV↗

arXiv2020

Structural phase transitions and photoluminescence mechanism in a layer of 3D hybrid perovskite nanocrystals

Although the structural phase transitions in single-crystal hybrid methyl-ammonium (MA) lead halide perovskites (MAPbX3, X = Cl, Br, I) are common phenomena, they have never been observed in the corresponding nanocrystals. Here we demonstrate that two-photon-excited photoluminescence (PL) spectroscopy is capable of monitoring the structural phase transitions in MAPbX3 nanocrystals because nonlinear susceptibilities govern the light absorption rates. We provide experimental evidence that the orthorhombic-to-tetragonal structural phase transition in a single layer of 20-nm-sized 3D MAPbBr3 nanocrystals is spread out within the 70 - 140 K range. This structural phase instability range arises because, unlike in single-crystal MAPbX3, free rotations of MA ions in the corresponding nanocrystals are no longer restricted by a long-range MA dipole order. The resulting configurational entropy loss can be even enhanced by the interfacial electric field arising due to charge separation at the MAPbBr3/ZnO heterointerface, extending the orthorhombic-to-tetragonal structural phase instability range from 70 to 230 K. We conclude that the weak sensitivity of conventional one-photon-excited PL spectroscopy to the structural phase transitions in 3D MAPbX3 nanocrystals results from the structural phase instability providing negligible distortions of PbX6 octahedra. In contrast, the intensity of two-photon-excited PL and electric-field-induced one-photon-excited PL still remains sensitive enough to weak structural distortions due to the higher rank tensor nature of nonlinear susceptibilities involved. We also show that room-temperature PL originates from the radiative recombination of the optical-phonon vibrationally excited polaronic quasiparticles with energies might exceed the ground-state Frohlich polaron and Rashba energies due to optical-phonon bottleneck.

physics.app-ph↗

arXiv2020

Distinguishing between dynamical and static Rashba effects in hybrid perovskite nanocrystals using transient absorption spectroscopy

The dynamical and static Rashba effects in hybrid methylammonium (MA) lead halide perovskites have recently been theoretically predicted. However, only the static effect was experimentally confirmed so far. Here we report on the dynamical Rashba effect observed using snapshot transient absorption spectral imaging with 400 nm pumping for a fully encapsulated film of 20-nm-sized 3D MAPbBr3 nanocrystals. The effect causes a 240 meV splitting of the lowest-energy absorption bleaching band, initially appearing over sub-ps timescale and progressively stabilizing to 60 meV during 500 ps. The integrated intensities of the split subbands demonstrate a photon-helicity-dependent asymmetry, thus proving the Rashba-type splitting and providing direct experimental evidence for the Rashba spin-split edge states in lead halide perovskite materials. The ultrafast dynamics is governed by the relaxation of two-photon-excited electrons in the Rashba spin-split system caused by a built-in electric field originating from dynamical charge separation in the entire MAPbBr3 nanocrystal.

physics.app-ph↗

arXiv2019

State transitions in the Morris-Lecar model under stable Lévy noise

This paper considers the state transition of the stochastic Morris-Lecar neuronal model driven by symmetric $α$-stable Lévy noise. The considered system is bistable: a stable fixed point (resting state) and a stable limit cycle (oscillating state), and there is an unstable limit cycle (borderline state) between them. Small disturbances may cause a transition between the two stable states, thus a deterministic quantity, namely the maximal likely trajectory, is used to analyze the transition phenomena in non-Gaussian stochastic environment. According to the numerical experiment, we find that smaller jumps of the Lévy motion and smaller noise intensity can promote such transition from the sustained oscillating state to the resting state. It also can be seen that larger jumps of the Lévy motion and higher noise intensity are conducive for the transition from the borderline state to the sustained oscillating state. As a comparison, Brownian motion is also taken into account. The results show that whether it is the oscillating state or the borderline state, the system disturbed by Brownian motion will be transferred to the resting state under the selected noise intensity.

math.DS↗

arXiv2019

Lévy noise induced escape in the Morris-Lecar model

The phenomenon of an excitable system producing a pulse under external or internal stimulation may be interpreted as a stochastic escape problem. This work addresses this issue by examining the Morris-Lecar neural model driven by symmetric α-stable Lévy motion. Two deterministic indices: the first escape probability and the mean first exit time, are adopted to analyse the state transition in this stochastic model. We calculate both indices in order to understand the transition from the escape region to the target region, and the area of higher indices in escape region. Additionally, we consider the special case of (Gaussian) Brownian motion to compare with (non-Gaussian) Lévy motion case. Our main results indicate that higher first escape probability promotes the transition, while the mean first exit time reflects the stability of the rest state with the selected escape region. The higher non-Gaussianity index and relatively small noise intensity are more prone to produce spikes. Moreover, by calculating both deterministic indices as functions of noise intensity ratio and non-Gaussianity index, we find that the effect of ion channel noise is more pronounced on the stochastic Morris-Lecar model than noise in the current. This work provides some mathematical understanding about the impact of non-Gaussian, heavy-tailed, burst-like fluctuations on excitable systems such as the Morris-Lecar system.

math.DS↗

arXiv2018

A role of random slow manifolds in detecting stochastic bifurcation

We consider the relation for the stochastic equilibrium states between the reduced system on a random slow manifold and the original system. This provides a theoretical basis for the reduction about sophisti- cated detailed models by the random slow manifold without significant damage to the overall qualitative properties. Based on this result, we reveal a role of random slow manifolds in detecting stochastic bifurca- tion by an example. The example exhibits a stochastic bifurcation phenomenon and possesses a random slow manifold that carries the stochastic bifurcation information of the system. Specifically, the lower dimensional reduced system on the random slow manifold retains the stochastic bifurcation phenomenon from the original system.

math.DS↗

arXiv2013

Regularized Discriminant Embedding for Visual Descriptor Learning

Images can vary according to changes in viewpoint, resolution, noise, and illumination. In this paper, we aim to learn representations for an image, which are robust to wide changes in such environmental conditions, using training pairs of matching and non-matching local image patches that are collected under various environmental conditions. We present a regularized discriminant analysis that emphasizes two challenging categories among the given training pairs: (1) matching, but far apart pairs and (2) non-matching, but close pairs in the original feature space (e.g., SIFT feature space). Compared to existing work on metric learning and discriminant analysis, our method can better distinguish relevant images from irrelevant, but look-alike images.

cs.CV↗