arXiv subjects

Fei Yu

Publications and source records attributed to Fei Yu.

At least 109 records · Page 6Linked to original sources

arXiv2023

Generation and characterization of frequency tuneable sub-15 fs pulses in a gas-filled hollow-core fiber pumped by a Yb:KGW laser

We investigate soliton self-compression and photoionization effects in an argon-filled antiresonant hollow-core photonic crystal fiber pumped with a commercial Yb:KGW laser. Before the onset of photoionization, we demonstrate self-compression of our 220~fs pump laser to 13~fs in a single and compact stage. By using the plasma driven soliton self-frequency blueshift, we also demonstrate a tuneable source from 1030 to ~700 nm. We fully characterize the compressed pulses using sum-frequency generation time-domain ptychography, experimentally revealing the full time-frequency plasma-soliton dynamics in hollow-core fibre for the first time.

physics.optics↗

arXiv2023

Phoenix: Democratizing ChatGPT across Languages

This paper presents our efforts to democratize ChatGPT across language. We release a large language model "Phoenix", achieving competitive performance among open-source English and Chinese models while excelling in languages with limited resources (covering both Latin and non-Latin languages). We believe this work will be beneficial to make ChatGPT more accessible, especially in countries where people cannot use ChatGPT due to restrictions from OpenAI or local goverments. Our data, code, and models are available at https://github.com/FreedomIntelligence/LLMZoo.

cs.CL↗

arXiv2023

Chiral Topological superconductivity in the OAI/SC/FMI heterostructure avoiding the subband problem

Implementing topological superconductivity (TSC) and Majorana states (MSs) is one of the most significant and challenging tasks in both fundamental physics and topological quantum computations. In this work, taking the obstructed atomic insulator (OAI) Nb3Br8, s-wave superconductor (SC) NbSe2 and ferromagnetic insulator (FMI) as example, we propose a new setup to realize the 2D chiral TSC and MSs in the OAI/SC/FMI heterostructure, which could avoid the subband problem effectively and has the advantage of huge Rashba spin-orbit coupling. As a result, the TSC phase can be stabilized in a wide region of chemical potential and Zeeman field, and four distinct TSC phases with superconducting Chern number N= -1, -2, -3, 3 can be achieved. Moreover, a 2D BdG Hamiltonian based on the triangular lattice of obstructed Wannier charge centers, combined with the s-wave superconductivity paring and Zeeman field, is constructed to understand the whole topological phase diagram analytically. These results expand the application of OAIs and pave a new way to realize the TSC and MSs with unique advantages.

cond-mat.mtrl-sci↗

arXiv2022

Measurements of microjoule-level, few-femtosecond ultraviolet dispersive-wave pulses generated in gas-filled hollow capillary fibers

High-energy ultraviolet pulse generation in gas-filled hollow capillary fibers (HCFs) through dispersive-wave-emission process, has attracted considerable attentions in recent several years due to its great application potentials in ultraviolet spectroscopy and photochemistry. While the ability of this technique to deliver high-energy, ultrafast ultraviolet pulses is widely recognized, few-fs duration of μJ-level dispersive-wave (DW) pulses has, however, escaped direct measurements. In this letter, we demonstrate that using several chirped mirrors, few-fs ultraviolet DW pulses can be obtained in atmosphere environment without the use of vacuum chambers. The pulse temporal profiles, measured using a self-diffraction frequency-resolved optical gating set-up, exhibit full-width-half-maximum pulse widths of 9.6 fs at 384 nm and 9.4 fs at 430 nm, close to the Fourier-transform limits. Moreover, theoretical and experimental studies reveal the strong influences of driving pulse energy and HCF length on temporal width and shape of the measured DW pulses. The ultraviolet pulses obtained in atmosphere environment with μJ-level pulse energy, few-fs pulse width and broadband wavelength tunability are ready to be used in many applications.

physics.optics↗

arXiv2022

Region-Aware Metric Learning for Open World Semantic Segmentation via Meta-Channel Aggregation

As one of the most challenging and practical segmentation tasks, open-world semantic segmentation requires the model to segment the anomaly regions in the images and incrementally learn to segment out-of-distribution (OOD) objects, especially under a few-shot condition. The current state-of-the-art (SOTA) method, Deep Metric Learning Network (DMLNet), relies on pixel-level metric learning, with which the identification of similar regions having different semantics is difficult. Therefore, we propose a method called region-aware metric learning (RAML), which first separates the regions of the images and generates region-aware features for further metric learning. RAML improves the integrity of the segmented anomaly regions. Moreover, we propose a novel meta-channel aggregation (MCA) module to further separate anomaly regions, forming high-quality sub-region candidates and thereby improving the model performance for OOD objects. To evaluate the proposed RAML, we have conducted extensive experiments and ablation studies on Lost And Found and Road Anomaly datasets for anomaly segmentation and the CityScapes dataset for incremental few-shot learning. The results show that the proposed RAML achieves SOTA performance in both stages of open world segmentation. Our code and appendix are available at https://github.com/czifan/RAML.

cs.CV↗

arXiv2021

Unsupervised Domain Adaptation in Semantic Segmentation Based on Pixel Alignment and Self-Training

This paper proposes an unsupervised cross-modality domain adaptation approach based on pixel alignment and self-training. Pixel alignment transfers ceT1 scans to hrT2 modality, helping to reduce domain shift in the training segmentation model. Self-training adapts the decision boundary of the segmentation network to fit the distribution of hrT2 scans. Experiment results show that PAST has outperformed the non-UDA baseline significantly, and it received rank-2 on CrossMoDA validation phase Leaderboard with a mean Dice score of 0.8395.

cs.CV↗

arXiv2021

Temperature-Dependent Group Delay of Photonic-Bandgap Hollow-Core Fiber Tuned by Surface-Mode Coupling

Surface modes (SM) are highly spatially localized modes existing at the core-cladding interface of photonic-bandgap hollow-core fiber (PBG-HCF). When coupling with SM, the air modes (AM) in the core would suffer a higher loss despite being spectrally within the cladding photonic bandgap, and would be highly dispersive around the avoided crossing (anti-crossing) wavelength. In this paper, we numerically demonstrate that such avoided crossings can play an important role in the tuning of the temperature dependence of group delay of AM of PBG-HCF. At higher temperatures, both the thermal-optic effect and thermal expansion contribute to the redshift of avoided crossing wavelength, giving rise to a temperature dependence of the AM dispersion. Numerical simulations show that the redshift of avoided crossing can significantly tune the thermal coefficient of delay (TCD) of PBG-HCF from -400 ps/km/K to 400 ps/km/K, approximately -120 ppm/K to 120 ppm/K. In comparison with the known tuning mechanism by the thermal-induced redshift of photonic bandgap [Fokoua et al., Optica 4, 659, 2017], the tuning of TCD by SM coupling presents a much broader tuning range and higher efficiency. Our finding would provide a new route to design PBG-HCF for propagation time sensitive applications.

physics.optics↗

arXiv2021

BEFD: Boundary Enhancement and Feature Denoising for Vessel Segmentation

Blood vessel segmentation is crucial for many diagnostic and research applications. In recent years, CNN-based models have leaded to breakthroughs in the task of segmentation, however, such methods usually lose high-frequency information like object boundaries and subtle structures, which are vital to vessel segmentation. To tackle this issue, we propose Boundary Enhancement and Feature Denoising (BEFD) module to facilitate the network ability of extracting boundary information in semantic segmentation, which can be integrated into arbitrary encoder-decoder architecture in an end-to-end way. By introducing Sobel edge detector, the network is able to acquire additional edge prior, thus enhancing boundary in an unsupervised manner for medical image segmentation. In addition, we also utilize a denoising block to reduce the noise hidden in the low-level features. Experimental results on retinal vessel dataset and angiocarpy dataset demonstrate the superior performance of the new BEFD module.

eess.IV↗

arXiv2021

ERNIE-ViL: Knowledge Enhanced Vision-Language Representations Through Scene Graph

We propose a knowledge-enhanced approach, ERNIE-ViL, which incorporates structured knowledge obtained from scene graphs to learn joint representations of vision-language. ERNIE-ViL tries to build the detailed semantic connections (objects, attributes of objects and relationships between objects) across vision and language, which are essential to vision-language cross-modal tasks. Utilizing scene graphs of visual scenes, ERNIE-ViL constructs Scene Graph Prediction tasks, i.e., Object Prediction, Attribute Prediction and Relationship Prediction tasks in the pre-training phase. Specifically, these prediction tasks are implemented by predicting nodes of different types in the scene graph parsed from the sentence. Thus, ERNIE-ViL can learn the joint representations characterizing the alignments of the detailed semantics across vision and language. After pre-training on large scale image-text aligned datasets, we validate the effectiveness of ERNIE-ViL on 5 cross-modal downstream tasks. ERNIE-ViL achieves state-of-the-art performances on all these tasks and ranks the first place on the VCR leaderboard with an absolute improvement of 3.7%.

cs.CV↗

arXiv2021

Photoionization-induced broadband dispersive wave generated in an Ar-filled hollow-core photonic crystal fiber

The resonance band in hollow-core photonic crystal fiber (HC-PCF), while leading to high-loss region in the fiber transmission spectrum, has been successfully used for generating phase-matched dispersive wave (DW). Here, we report that the spectral width of the resonance-induced DW can be largely broadened due to plasma-driven blueshifting soliton. In the experiment, we observed that in a short length of Ar-filled single-ring HC-PCF the soliton self-compression and photoionization effects caused a strong spectral blueshift of the pump pulse, changing the phase-matching condition of the DW emission process. Therefore, broadening of DW spectrum to the longer-wavelength side was obtained with several spectral peaks, which correspond to the generation of DW at different positions along the fiber. In the simulation, we used super-Gauss windows with different central wavelengths to filter out these DW spectral peaks, and studied the time-domain characteristics of these peaks respectively using Fourier transform method. The simulation results verified that these multiple-peaks on the DW spectrum have different delays in the time domain, agreeing well with our theoretical prediction. Remarkably, we found that the whole time-domain DW trace can be compressed to ~29 fs using proper chirp compensation. The experimental and numerical results reported here provide some insight into the resonance-induced DW generation process in gas-filled HC-PCFs, they could also pave the way to ultrafast pulse generation using DW-emission mechanism.

physics.optics↗

arXiv2020

Photoionization-assisted, high-efficiency emission of dispersive wave in gas-filled hollow-core photonic crystal fibers

We demonstrate that the phase-matched dispersive wave (DW) emission within the resonance band of a 25-cm-long gas-filled hollow-core photonic crystal fiber (HC-PCF) can be strongly enhanced by the photoionization effect of the pump pulse. In the experiments we observe that as the pulse energy increases, the pump pulse gradually shifts to shorter wavelengths due to soliton-plasma interactions. When the central wavelength of the blueshifting soliton is close to the resonance band of the HC-PCF, high-efficiency energy transfer from the pump light to the DW in the visible region can be obtained. During this DW emission process, we also observe that the spectral center of the DW gradually shifts to longer wavelengths leading to a slightly-increased DW bandwidth, which can be well explained as the consequence of phase-matched coupling between the pump pulse and the DW. In particular, at an input pulse energy of 6 uJ, the spectral ratio of the DW at the fiber output is measured to be as high as ~53% together with a conversion efficiency of ~19%. These experimental results, explained by numerical simulations, pave the way to high-brightness light sources based on high-efficiency frequency-upconversion processes in gas-filled HC-PCFs.

physics.optics↗

arXiv2019

Higher-order accurate diffuse-domain methods for partial differential equations with Dirichlet boundary conditions in complex, evolving geometries

The diffuse-domain, or smoothed boundary, method is an attractive approach for solving partial differential equations in complex geometries because of its simplicity and flexibility. In this method the complex geometry is embedded into a larger, regular domain. The original PDE is reformulated using a smoothed characteristic function of the complex domain and source terms are introduced to approximate the boundary conditions. The reformulated equation, which is independent of the dimension and domain geometry, can be solved by standard numerical methods and the same solver can be used for any domain geometry. A challenge is making the method higher-order accurate. For Dirichlet boundary conditions, which we focus on here, current implementations demonstrate a wide range in their accuracy but generally the methods yield at best first order accuracy in $ε$, the parameter that characterizes the width of the region over which the characteristic function is smoothed. Typically, $ε\propto h$, the grid size. Here, we analyze the diffuse-domain PDEs using matched asymptotic expansions and explain the observed behaviors. Our analysis also identifies simple modifications to the diffuse-domain PDEs that yield higher-order accuracy in $ε$, e.g., $O(ε^2)$ in the $L^2$ norm and $O(ε^p)$ with $1.5\le p\le 2$ in the $L^{\infty}$ norm. Our analytic results are confirmed numerically in stationary and moving domains where the level set method is used to capture the dynamics of the domain boundary and to construct the smoothed characteristic function.

math.NA↗

arXiv2019

Attentive Geo-Social Group Recommendation

Social activities play an important role in people's daily life since they interact. For recommendations based on social activities, it is vital to have not only the activity information but also individuals' social relations. Thanks to the geo-social networks and widespread use of location-aware mobile devices, massive geo-social data is now readily available for exploitation by the recommendation system. In this paper, a novel group recommendation method, called attentive geo-social group recommendation, is proposed to recommend the target user with both activity locations and a group of users that may join the activities. We present an attention mechanism to model the influence of the target user $u_T$ in candidate user groups that satisfy the social constraints. It helps to retrieve the optimal user group and activity topic candidates, as well as explains the group decision-making process. Once the user group and topics are retrieved, a novel efficient spatial query algorithm SPA-DF is employed to determine the activity location under the constraints of the given user group and activity topic candidates. The proposed method is evaluated in real-world datasets and the experimental results show that the proposed model significantly outperforms baseline methods.

cs.SI↗

arXiv2019

PGU-net+: Progressive Growing of U-net+ for Automated Cervical Nuclei Segmentation

Automated cervical nucleus segmentation based on deep learning can effectively improve the quantitative analysis of cervical cancer. However, accurate nuclei segmentation is still challenging. The classic U-net has not achieved satisfactory results on this task, because it mixes the information of different scales that affect each other, which limits the segmentation accuracy of the model. To solve this problem, we propose a progressive growing U-net (PGU-net+) model, which uses two paradigms to extract image features at different scales in a more independent way. First, we add residual modules between different scales of U-net, which enforces the model to learn the approximate shape of the annotation in the coarser scale, and to learn the residual between the annotation and the approximate shape in the finer scale. Second, we start to train the model with the coarsest part and then progressively add finer part to the training until the full model is included. When we train a finer part, we will reduce the learning rate of the previous coarser part, which further ensures that the model independently extracts information from different scales. We conduct several comparative experiments on the Herlev dataset. The experimental results show that the PGU-net+ has superior accuracy than the previous state-of-the-art methods on cervical nuclei segmentation.

eess.IV↗

arXiv2019

Multi-level Domain Adaptive learning for Cross-Domain Detection

In recent years, object detection has shown impressive results using supervised deep learning, but it remains challenging in a cross-domain environment. The variations of illumination, style, scale, and appearance in different domains can seriously affect the performance of detection models. Previous works use adversarial training to align global features across the domain shift and to achieve image information transfer. However, such methods do not effectively match the distribution of local features, resulting in limited improvement in cross-domain object detection. To solve this problem, we propose a multi-level domain adaptive model to simultaneously align the distributions of local-level features and global-level features. We evaluate our method with multiple experiments, including adverse weather adaptation, synthetic data adaptation, and cross camera adaptation. In most object categories, the proposed method achieves superior performance against state-of-the-art techniques, which demonstrates the effectiveness and robustness of our method.

cs.CV↗

arXiv2019

Annotation-Free Cardiac Vessel Segmentation via Knowledge Transfer from Retinal Images

Segmenting coronary arteries is challenging, as classic unsupervised methods fail to produce satisfactory results and modern supervised learning (deep learning) requires manual annotation which is often time-consuming and can some time be infeasible. To solve this problem, we propose a knowledge transfer based shape-consistent generative adversarial network (SC-GAN), which is an annotation-free approach that uses the knowledge from publicly available annotated fundus dataset to segment coronary arteries. The proposed network is trained in an end-to-end fashion, generating and segmenting synthetic images that maintain the background of coronary angiography and preserve the vascular structures of retinal vessels and coronary arteries. We train and evaluate the proposed model on a dataset of 1092 digital subtraction angiography images, and experiments demonstrate the supreme accuracy of the proposed method on coronary arteries segmentation.

eess.IV↗

arXiv2018

Exploring interacting holographic dark energy in a perturbed universe with parameterized post-Friedmann approach

The model of holographic dark energy in which dark energy interacts with dark matter is investigated in this paper. In particular, we consider the interacting holographic dark energy model in the context of a perturbed universe, which was never investigated in the literature. To avoid the large-scale instability problem in the interacting dark energy cosmology, we employ the generalized version of the parameterized post-Friedmann approach to treat the dark energy perturbations in the model. We use the current observational data to constrain the model. Since the cosmological perturbations are considered in the model, we can then employ the redshift-space distortions (RSD) measurements to constrain the model, in addition to the use of the measurements of expansion history, which was either never done in the literature. We find that, for both the cases with $Q=βHρ_{\rm c}$ and $Q=βH_0ρ_{\rm c}$, the interacting holographic dark energy model is more favored by the current data, compared to the holographic dark energy model without interaction. It is also found that, with the help of the RSD data, a positive coupling $β$ can be detected at the $2.95σ$ statistical significance for the case of $Q=βH_0ρ_{\rm c}$.

astro-ph.CO↗

arXiv2018

High fidelity fibre-based physiological sensing deep in tissue

Physiological sensing deep in tissue, remains a clinical challenge. Here a flexible miniaturised sensing optrode providing a platform to perform minimally invasive in vivo in situ measurements is reported. Silica microspheres covalently coupled with a high density of ratiometrically configured fluorophores were deposited into etched pits on the distal end of a 150 micron diameter multicore optical fibre. With this platform, multiplexed photonic measurements of pH and oxygen concentration with high precision in the distal alveolar space of the lung is reported. We demonstrated the phenomenon that high-density deposition of carboxyfluorescein covalently coupled to silica microspheres shows an inverse shift in fluorescence in response to varying pH. This platform delivered fast and accurate measurements, near instantaneous response time, no photobleaching, immunity to power fluctuations and a flexible architecture for addition of multiple sensors.

physics.med-ph↗