A5101583457- Photonic and Optical Devices
- Photonic Crystals and Applications
- Natural Language Processing Techniques
- Topic Modeling
- Peer-to-Peer Network Technologies
- Semiconductor Lasers and Optical Devices
- Caching and Content Delivery
- Optical Coatings and Gratings
- Cloud Computing and Resource Management
- Second Language Acquisition and Learning
- Advanced Photonic Communication Systems
- Advanced Data Storage Technologies
- HVDC Systems and Fault Protection
- Advanced Fiber Laser Technologies
- Optical Network Technologies
- Parallel Computing and Optimization Techniques
- Acute Myocardial Infarction Research
- Graph Theory and Algorithms
- Machine Learning in Healthcare
- Tissue Engineering and Regenerative Medicine
- Distributed systems and fault tolerance
- Artificial Intelligence in Healthcare
- Cardiovascular Function and Risk Factors
- Power System Optimization and Stability
- Electrospun Nanofibers in Biomedical Applications
Baoji City Central Hospital
2024
Xi'an Technological University
2023
University of Southampton
2016-2021
Peking University
2011-2016
We present several fundamental photonic building blocks based on suspended silicon waveguides supported by a lateral cladding comprising subwavelength grating metamaterial.We discuss the design, fabrication, and characterization of waveguide bends, multimode interference devices Mach-Zehnder interferometers for 3715 -3800 nm wavelength range, demonstrated first time in this platform.The propagation loss 0.82 dB/cm is reported, some lowest yet achieved range.These results establish direct...
In this Letter, we report suspended silicon waveguides operating at a wavelength of 7.67 μm with propagation loss 3.1 0.3 dB∕cm.To our knowledge, is the first demonstration low-loss such long wavelength, comparable to other platforms that use more exotic materials.The Si waveguide core supported by sub-wavelength grating provides lateral optical confinement while also allowing access buried oxide layer so it can be wet etched using hydrofluoric acid.We demonstrate bends and s-bends.
Germanium is a material of high interest for midinfrared (MIR) integrated photonics due to its CMOS compatibility and wide transparency window covering the 2-15 µm spectral region exceeding 4 8 limit Silicon-on-Insulator (SOI) platform Si respectively.In this Letter, we report suspended germanium waveguides operating at wavelength 7.67 with propagation loss 2.6 ± 0.3 dB/cm.To our knowledge, first demonstration low-loss such long wavelength.Suspension waveguide achieved by defining holes...
Graph processing systems have been widely used in enterprises like online social networks to process their daily jobs. With the fast growing of applications, they efficiently handle massive concurrent However, due inherent design for single job, existing incur great inefficiency memory use and fault tolerance. Motivated by this, this paper we introduce Seraph, a graph system that enables efficient job-level parallelism. Seraph is designed based on decoupled data model, which allows multiple...
In this paper we review our recent results on group IV mid-infrared photonic devices.In particular, passive structures suitable for long wavelength operation, such as suspended Si, Ge-on-Si and Ge, are analyzed.In addition, Ge-on-insulator waveguides have been characterized at 3.8 μm.Several active devices also realized: optical modulators in silicon germanium, graphene detectors operating shorter mid-IR wavelengths.
The multilingual neural machine translation (MNMT) enables arbitrary translations across multiple languages by training a model with limited parameters using parallel data only. However, the performance of such MNMT models still lags behind that large language (LLMs), limiting their practicality. In this work, we address limitation introducing registering to achieve new state-of-the-art decoder-only models. Specifically, insert set artificial tokens specifying target language, called...
Abstract Drug-induced cardiac toxicity is a critical concern in drug development, often leading to unreliable results and potential withdrawal due the lack of mature cardiomyocytes screening assays. This study aimed enhance cardiomyocyte maturation by combining structural conductive stimuli, specifically using Poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) coated PDMS membrane mechanical stimulation (MS) system. A 10% tensile strain at frequency 1 Hz was applied via...
The heterogeneous nature of Peer-to-Peer (P2P) networks can be exploited to optimize a wide range applications. But this requires an accurate characterization peer heterogeneity, which is difficult due huge population and disparate properties individuals. To overcome barrier, we conduct thorough inspection heterogeneity in terms individual churn resource capacity. We make first attempt (i) experimentally reduce dynamics simple distributions with individual-specific parameters, (ii) provide...
Phrase-level dense retrieval has shown many appealing characteristics in downstream NLP tasks by leveraging the fine-grained information that phrases offer. In our work, we propose a new task formulation of retrieval, cross-lingual contextualized phrase which aims to augment applications addressing polysemy using context information. However, lack specific training data and models are primary challenges achieve goal. As result, extract pairs word alignment automatically induced from parallel...
A Knowledge Graph (KG) is the directed graphical representation of entities and relations in real world. KG can be applied diverse Natural Language Processing (NLP) tasks where knowledge required. The need to scale up complete automatically yields Embedding (KGE), a shallow machine learning model that suffering from memory training time consumption issues. To mitigate computational load, we propose parameter-sharing method, i.e., using conjugate parameters for complex numbers employed KGE...
Understanding representation transfer in multilingual neural machine translation can reveal the representational issue causing zero-shot deficiency. In this work, we introduce identity pair, a sentence translated into itself, to address lack of base measure investigations, as pair represents optimal state among any language transfers. our analysis, demonstrate that encoder transfers source subspace target instead language-agnostic state. Thus, deficiency arises because representations are...
Multilingual neural machine translation models support fine-tuning hundreds of languages simultaneously. However, on full parameters solely is inefficient potentially leading to negative interactions among languages. In this work, we demonstrate that the for a language occurs in its intrinsic language-specific subspace with tiny fraction entire parameters. Thus, propose LoRA isolate subspaces. Furthermore, architecture learning techniques and introduce gradual pruning schedule during...
Acute ST-segment elevation myocardial infarction (STEMI) remains a major contributor to morbidity and mortality worldwide. The no-reflow phenomenon following percutaneous coronary intervention (PCI) complicates the clinical outcome of STEMI. This study aimed identify valuable predictor for phenomenon.
Existing multilingual neural machine translation (MNMT) approaches mainly focus on improving models with the encoder-decoder architecture to translate multiple languages. However, decoder-only has been explored less in MNMT due its underperformance when trained parallel data solely. In this work, we attribute issue of lack language transfer capability. Specifically, is insufficient encoding source tokens target features. We propose dividing decoding process into two stages so that are...
Summary Raft is a consensus algorithm that implements highly available replicas. As raft requires leader to drive the protocol execution, it usually employs round‐robin election mechanism select random among all nodes. However, when elected suffers from performance issues, will temporarily lose aliveness. To address this issue, we propose multi‐strategy allows nodes proactively trigger new round of they believe and replace with one in designated priority queue. The formal analysis...
BACKGROUND There is increasing interest in clinical research with electronic medical data, but it often faces the challenges of heterogeneity between hospitals. Our objective was to develop a single numerical score for characterizing such via computing inpatient mortality treating acute myocardial infarction (AMI) patients based on diagnostic information recorded database Discharge Summary Reports (DSR). MATERIAL AND METHODS Using 4 216 135 DSRs 49 tertiary hospitals from 2006 2010 Beijing,...
Both silicon and germanium are very interesting materials for mid-infrared applications, particularly sensing, due to their low material losses at long wavelengths. Silicon is loss up 8 μm, 15 μm [1]. To fully exploit these wavelength transmission ranges of Si Ge, build Ge photonic circuits, cladding need have the same or larger than Ge. One solution remove a lossy substrate create suspended structures with air cladding. In case SOI, buried oxide can be removed by HF [2,3], Ge-on-Si...
Multilingual neural machine translation can translate unseen language pairs during training, i.e. zero-shot translation. However, the is always unstable. Although prior works attributed instability to domination of central language, e.g. English, we supplement this viewpoint with strict dependence non-centered languages. In work, propose a simple, lightweight yet effective language-specific modeling method by adapting languages and combining shared information counteract Experiments...
This paper presents the design, implementation, and application guidelines for several Master Power Plant Controllers (MPPCs) that have been used in project specific locations across parts of North America Australia. The MPPCs described here provide advanced grid functions hybrid power plants wind farms a mix different OEM technologies associated control strategies. Over course this paper, functionalities are site uncommon described. As dynamic models were not available simulation studies,...
Silicon and germanium are transparent up to approximately 8 µm 15 µm, respectively, thus offering a range of applications in biochemical environmental sensing, medicine, astronomy communications. In this paper we present our recent results on silicon modulators detectors suitable for mid-infrared
In this paper we report suspended silicon waveguides operating at a wavelength of 7.67 mm with propagation loss 3.1 ? 0.3 dB/cm. To our knowledge is the first demonstration low such long wavelength, comparable to other platforms that use more exotic materials. The Si waveguide core supported by subwavelength grating provides lateral optical confinement, while also allowing access buried oxide layer so it can be wet etched using hydrofluoric acid. We demonstrate bends and s-bends.
The mid-infrared (MIR) wavelength region contains unique absorption fingerprints of many molecules and substances. Therefore, the main application area in MIR can be sensing. Si Ge are transparent this range (Si up to 8 μm, 15 μm) due their excellent electrical optical properties CMOS compatibility they offer suitable platforms for realisation compact sensors. However, claddings used telecom usually not at longer wavelengths high absorption, so alternative solutions need investigated. In...