A5045547813- Advanced MIMO Systems Optimization
- Topic Modeling
- Natural Language Processing Techniques
- Video Surveillance and Tracking Methods
- Advanced Wireless Network Optimization
- Wireless Communication Networks Research
- Logic, Reasoning, and Knowledge
- Cooperative Communication and Network Coding
- Human Pose and Action Recognition
- Indoor and Outdoor Localization Technologies
- Energy Efficient Wireless Sensor Networks
- Advanced Neural Network Applications
- Advanced Vision and Imaging
- Multimodal Machine Learning Applications
- Embedded Systems and FPGA Design
- Advanced Algebra and Logic
- Anomaly Detection Techniques and Applications
- Logic, programming, and type systems
- Underwater Vehicles and Communication Systems
- Advanced Wireless Communication Techniques
- Wireless Sensor Networks and IoT
- Semantic Web and Ontologies
- Advanced Text Analysis Techniques
- IoT-based Smart Home Systems
- Robotics and Sensor-Based Localization
Tsinghua University
2003-2024
Beihang University
2010-2024
Zhengzhou University
2023
Baidu (China)
2016-2023
Shanghai Artificial Intelligence Laboratory
2023
Xiamen University
2023
CHN Energy (China)
2023
Baotou Medical College
2022
Southwest Jiaotong University
2022
Guangxi University for Nationalities
2022
Transfer learning, where a model is first pre-trained on data-rich task before being fine-tuned downstream task, has emerged as powerful technique in natural language processing (NLP). The effectiveness of transfer learning given rise to diversity approaches, methodology, and practice. In this paper, we explore the landscape techniques for NLP by introducing unified framework that converts all text-based problems into text-to-text format. Our systematic study compares pre-training...
Models for many natural language tasks benefit from the flexibility to use overlapping, non-independent features. For example, need labeled data can be drastically reduced by taking advantage of domain knowledge in form word lists, part-of-speech tags, character n-grams, and capitalization patterns. While it is difficult capture such inter-dependent features with a generative probabilistic model, conditionally-trained models, as conditional maximum entropy handle them well. There has been...
We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities is more compute-efficient than its predecessor PaLM. 2 Transformer-based trained using mixture of objectives. Through extensive evaluations on English language, tasks, we demonstrate significantly improved quality downstream tasks across different sizes, while simultaneously exhibiting faster efficient inference compared to This efficiency enables broader deployment also...
In this paper we propose a new type of random CSP model, called Model RB, which is revision to the standard B. It proved that phase transitions from region where almost all problems are satisfiable unsatisfiable do exist for RB as number variables approaches infinity. Moreover, critical values at occur also known exactly. By relating hardness B, it shown there lot hard instances in RB.
Chinese poetry generation is a very challenging task in natural language processing. In this paper, we propose novel two-stage generating method which first plans the sub-topics of poem according to user's writing intent, and then generates each line sequentially, using modified recurrent neural network encoder-decoder framework. The proposed planning-based can ensure that generated coherent semantically consistent with intent. A comprehensive evaluation human judgments demonstrates our...
Zixuan Li, Xiaolong Jin, Saiping Guan, Wei Jiafeng Guo, Yuanzhuo Wang, Xueqi Cheng. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.
In this paper, we present a new large-scale dataset for the video panoptic segmentation task, which aims to assign semantic classes and track identities all pixels in video. As ground truth task is difficult annotate, previous datasets are limited by either small scales or number of scenes. contrast, our VIdeo Panoptic Segmentation Wild (VIPSeg) provides 3,536 videos 84,750 frames with pixel-level annotations, covering wide range real-world scenarios categories. To best knowledge, VIPSeg...
This paper describes our application of conditional random fields with feature induction to a Hindi named entity recognition task. With only five days development time and little knowledge this language, we automatically discover relevant features by providing large array lexical tests using construct the that most increase likelihood. In an effort reduce overfitting, use combination Gaussian prior early stopping based on results 10-fold cross validation.
Femtocell is a small access point using the wire broadband connections or wireless technologies to mobile operator's network for user equipment(UE), which can provide better indoor coverage and satisfy upcoming demand of high data rate communication system.Femtocell related handover cost reduction one important targets in LTE-Advanced SON (Self-Organising Networks). In this paper, optimization algorithm based on UE's mobility state proposed. An analytical model was presented signalling...
While question answering (QA) with neural network, i.e. QA, has achieved promising results in recent years, lacking of large scale real-word QA dataset is still a challenge for developing and evaluating system. To alleviate this problem, we propose human annotated real-world WebQA more than 42k questions 556k evidences. As existing methods resolve either as sequence generation or classification/ranking they face challenges expensive softmax computation, unseen answers handling separate...
Shuming Ma, Xu Sun, Wei Li, Sujian Wenjie Xuancheng Ren. Proceedings of the 2018 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.
Recently, adversarial attacks have been applied in visual object tracking to deceive deep trackers by injecting imperceptible perturbations into video frames. However, previous work only generates the video-specific perturbations, which restricts its application scenarios. In addition, existing are difficult implement reality due real-time of and re-initialization mechanism. To address these issues, we propose an offline universal attack called Efficient Universal Shuffle Attack. It takes...
In this paper, we present a framework for active contour-based visual tracking using level sets. The main components of our include initialization, color-based contour evolution, adaptive shape-based evolution non-periodic motions, dynamic periodic and the handling abrupt motions. For initialization tracking, develop an optical flow-based algorithm automatically initializing contours at first frame. Markov random field theory is used to measure correlations between values neighboring pixels...
Femtocell networks have great potential for mobile applications. However, interference due to the co-existence of macrocells and femtocells is a serious problem. In addition, dense introduce severe inter-system interference. this paper, graph theory based dynamic sub-band allocation technique presented avoid downlink We model cells their mutual as elements, nodes weighted edges, respectively. To maintain tolerable level, total bandwidth divided into number sub-bands, these are assigned using...
Indoor wireless localization toward the next generation Wi-Fi access point has attracted considerable attention due to presentation of state-of-art fine time measurement (FTM) protocol. In order improve autonomy, accuracy, and universality positioning based on Internet Things (IoT) terminals, this article proposes a hybrid system which contains integration FTM, crowdsourced received signal strength indicator (RSSI) fingerprinting micro-electro-mechanical-system (MEMS) sensors (H-WPS). A...
Chiral γ-hydroxy-γ-lactams were versatile building blocks for the synthesis of many important bioactive compounds. Herein, we disclosed a regio- and enantioselective direct α-selective allylation amide moiety maleimides with...
The northwestern region of China is characterized by loess soil and seasonal permafrost. Due to the combined effects its unique climate precipitation patterns, local roads frequently suffer from issues such as foundation settlement, erosion, collapse, which pose significant risks both road construction safe operation. This study examines a typical high subgrade in Northwest China, where scaled laboratory model experiment was conducted. research investigates impact water infiltration at slope...
Natural Language Generation (NLG) has made great progress in recent years due to the development of deep learning techniques such as pre-trained language models. This advancement resulted more fluent, coherent and even properties controllable (e.g. stylistic, sentiment, length etc.) generation, naturally leading downstream tasks abstractive summarization, dialogue machine translation, data-to-text generation. However, faithfulness problem that generated text usually contains unfaithful or...
In this paper, spectrum holes in space-time dimensions are exploited to improve the utilization of cognitive radio networks (CRNs). To achieve goal, we design architecture three regions, which separately known as black region, gray and white region. Temporal exist where secondary users (SUs) can opportunistically access licensed with interweave sharing. SUs exploit spatial opportunities transmit at any time by taking advantage their long distances from primary (PUs), without causing severe...