A5100344431- Topic Modeling
- Natural Language Processing Techniques
- Sentiment Analysis and Opinion Mining
- Advanced Text Analysis Techniques
- Multimodal Machine Learning Applications
- Text and Document Classification Technologies
- Alzheimer's disease research and treatments
- Polymer crystallization and properties
- Energy Load and Power Forecasting
- Educational and Technological Research
- Domain Adaptation and Few-Shot Learning
- Complex Systems and Time Series Analysis
- Statistical and Computational Modeling
- Morphological variations and asymmetry
- Amyotrophic Lateral Sclerosis Research
- Hearing, Cochlea, Tinnitus, Genetics
- Speech Recognition and Synthesis
- Visual perception and processing mechanisms
- Hybrid Renewable Energy Systems
- Parkinson's Disease Mechanisms and Treatments
- Advanced Image and Video Retrieval Techniques
- Rheology and Fluid Dynamics Studies
- Synthetic Organic Chemistry Methods
- Speech and dialogue systems
- Traditional and Medicinal Uses of Annonaceae
Peking University
2014-2024
Weihai Maternal and Child Health Hospital
2022
Weihai Municipal Hospital
2022
Qingdao University
2010-2022
Shanghai University of Finance and Economics
2022
Chongqing University
2022
Amazon (Germany)
2021
Institute of Computing Technology
2020
Chinese Academy of Sciences
2017-2020
Guangdong University of Technology
2020
Unsupervised neural machine translation (NMT) is a recently proposed approach for which aims to train the model without using any labeled data. The models unsupervised NMT often use only one shared encoder map pairs of sentences from different languages shared-latent space, weak in keeping unique and internal characteristics each language, such as style, terminology, sentence structure. To address this issue, we introduce an extension by utilizing two independent encoders but sharing some...
Abstract A type II intramolecular oxidopyrylium‐mediated [5+2] cycloaddition reaction allows the efficient and diastereoselective formation of various highly functionalized synthetically challenging bridged seven‐membered ring systems (such as bicyclo[4.4.1]undecane, bicyclo[4.3.1]decane, bicyclo[5.4.1]dodecane, bicyclo[6.4.1]]tridecane). This simple, thermal, direct transformation has a broad substrate scope is high yielding, with functional‐group tolerance unique endo selectivity. The...
Stock forecast is a crucial yet challenging task in modern quantitative trading. Given theoretical and investment merits, recently variety of deep learning methods have been proposed for automatically simulating stock movements from historical time series. However, these typically follow the i.i.d. assumption that actually contradicts complex trading environment. In reality, individual stocks often exhibit diverse volatility patterns, while macro market scenarios may also change over time,...
In many practical applications, neural machine translation systems have to deal with the input from automatic speech recognition (ASR) which may contain a certain number of errors. This leads two problems degrade performance. One is discrepancy between training and testing data other error caused by errors ruin whole translation. this paper, we propose method handle so as generate robust ASR First, simulate in that distribution test consistent. Second, focus on homophone words similar...
Liqiang Xiao, Jun Ma, Xin Luna Dong, Pascual Martínez-Gómez, Nasser Zalmout, Wei Chen, Tong Zhao, Hao He, Yaohui Jin. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021.
Learning semantic sentence embeddings is beneficial to a variety of natural language processing tasks. Recently, methods using the contrastive learning framework fine-tune pre-trained models have been proposed and achieved significant performance on embeddings. However, are easy “overfit” goal. With training learning, gap between test tasks leads unstable even declining For this reason, existing rely labeled development set frequently evaluate get best checkpoints. In such way, limited when...
Cross-domain Consultation Systems have become essential in numerous critical applications, for instance, an online citizen complaint system. However, addressing complaints with distinct orality characteristics often necessitates retrieving and integrating knowledge from diverse professional domains. This scenario represents a typical cross-domain problem. Nevertheless, the prevailing approach of utilizing generative large language models to tackle this problem presents challenges including...
Multi-label few-shot image recognition aims to identify multiple unseen objects using only a handful of examples. Recent methods typically tune pre-trained vision-language models with shared or class-specific prompts. However, they still have drawbacks. Tuning prompt is insufficient for all samples especially when the tasks are complex and tuning specific prompts each class inevitable lose generalization ability, thus failing capture diverse visual knowledge. To address these issues, we...
Motivation: Most fMRI column studies don’t consider the potentially different laminar responses along cortical depth. Goal(s): In this work, we demonstrate feasibility of mapping ocular dominance columns (ODCs) at multiple depths in cat visual cortex using CBV-weighted fMRI. Approach: To enable monocular input to either eye a cat, customized goggle that can avoid light leakage and switch eyes easily. Results: By employing mesoscopic ultrahigh magnetic field (9.4 Tesla), observed...
Generating qualitative responses has always been a challenge for human-computer dialogue systems. Existing systems generally derive from either retrieval-based or generative-based approaches, both of which have their own pros and cons. Despite the natural idea an ensemble model two, existing methods only focused on leveraging one approach to enhance another, we argue however that they can be further mutually enhanced with proper training strategy. In this paper, propose ensembleGAN,...
Domain adaptation on time series data is an important but challenging task. Most of the existing works in this area are based learning domain-invariant representation with help restrictions like MMD. However, such extraction a non-trivial task for data, due to complex dependence among timestamps. In detail, fully dependent series, small change lags or offsets may lead difficulty domain invariant extraction. Fortunately, stability causality inspired us explore structure data. To reduce...
Introduction At least 60% of cases severe hearing loss result from genetic factors. In this study screening was carried out for common deafness in women childbearing age to prevent and birth defects via providing counseling follow-up services high-risk families. Material methods total 60,391 pre-pregnancy/early-gestation who received treatment second-level or above hospitals Weihai February 2017 December 2019 were selected. Venous peripheral blood collected make dried slices on filter paper...