A5101477342- Cooperative Communication and Network Coding
- Mobile Ad Hoc Networks
- Natural Language Processing Techniques
- Opportunistic and Delay-Tolerant Networks
- Topic Modeling
- Vehicular Ad Hoc Networks (VANETs)
- Wireless Communication Security Techniques
- Wireless Networks and Protocols
- Energy Efficient Wireless Sensor Networks
- Full-Duplex Wireless Communications
- Cellular Automata and Applications
- Interconnection Networks and Systems
- Image and Signal Denoising Methods
- Advanced Data Storage Technologies
- Error Correcting Code Techniques
- Advanced Wireless Communication Technologies
- Robotics and Automated Systems
- Network Security and Intrusion Detection
- Intelligent Tutoring Systems and Adaptive Learning
- Internet Traffic Analysis and Secure E-voting
- Digital Media and Visual Art
- Mobile Agent-Based Network Management
- Coding theory and cryptography
- Cybersecurity and Information Systems
- Image Enhancement Techniques
Institute of Computing Technology
2024
Chinese Academy of Sciences
2024
Jiaying University
2021-2024
Qingdao Municipal Center for Disease Control and Prevention
2023
Shandong Institute of Commerce & Technology
2022
Bohai University
2010-2019
Xi'an University of Technology
2019
Dalian Minzu University
2008-2016
Dalian University of Technology
2016
University of Regina
2015
Best-of-N (BoN) sampling, a common strategy for test-time scaling of Large Language Models (LLMs), relies on reward models to select the best candidate solution from multiple generations. However, traditional often assign arbitrary and inconsistent scores, limiting their effectiveness. To address this, we propose Pairwise Reward Model (Pairwise RM) combined with knockout tournament BoN sampling. Instead assigning absolute given one math problem, RM evaluates two solutions' correctness...
In this paper, we propose KnowCoder, a Large Language Model (LLM) to conduct Universal Information Extraction (UIE) via code generation. KnowCoder aims develop kind of unified schema representation that LLMs can easily understand and an effective learning framework encourages follow schemas extract structured knowledge accurately. To achieve these, introduces code-style method uniformly transform different into Python classes, with which complex information, such as constraints among tasks...
Background: Severe acute respiratory syndrome coronavirus (SARS-CoVs) have emerged as a global health threat, which had caused high rate of mortality. There is an urgent need to find effective drugs against these viruses. Objective: This study aims predict the activity unsymmetrical aromatic disulfides by constructing QSAR model, and design new compounds according structural physicochemical attributes responsible for higher towards SARS-CoVs main protease. Methods: All molecules were...
This paper analyzed the security of constant dimensional subspace code against wiretap attacks. The was measured in probability with which an eavesdropper guessed source message successfully. With methods linear algebra and combinatorics, analytic solution obtained. Performance compared to several secure network coding schemes from perspective security, flexibility, complexity, independence, etc. comparison showed did not have perfect but it achieved probabilistic low complexity. As a...
Because it is hard to depict real world traces, researchers proposed some mobility models in the study of mobile ad-hoc networks. It aware recently that will lead non-uniform node distribution, which decreases credibility and validity simulations. Many reconsider distribution characteristic for random waypoint model have gotten valuable quantitative results. But similar investigations on direction (RD) square region are still missing. We presented a detailed analysis probability behavior RD...
Aiming at the inefficiency of polynomial time algorithm linear network coding (NC) for multicast applications, authors propose an improved which combines with subtree decomposition. The is composed five steps, including line graph transforming, decomposition, static set generating, assigning global vector, and calculation local vector. Subtree decomposition decreases scale complexity steeply, so it efficient NC multicast. An example given to illustrate implementation algorithm. Detailed...
Rate region precisely measures the transmission ability of a network. Three graph processing techniques are presented: reduced graph, union and virtual node. Applying these techniques, an operable method to completely solve calculation rate regions for unicast routing networks is developed. The correctness proved theoretically illustrated with example. This helpful in both theoretical perspective industrial implementation.
Learning an image filter from a pair of images composed original and its filtered version has attracted increasing attention in the processing region. However, existing methods only take color information into consideration to learn filter, which cannot accurately depict luminance some style changes so that generated learning results are inaccurate. This paper proposes Laplace operator based multi-channel approach version. Different previous method, we adopt strategy can reconstruct image....
Modern Large Language Models (LLMs) have showcased remarkable prowess in various tasks necessitating sophisticated cognitive behaviors. Nevertheless, a paradoxical performance discrepancy is observed, where these models underperform seemingly elementary like relation extraction and event due to two issues conventional evaluation. (1) The imprecision of existing evaluation metrics that struggle effectively gauge semantic consistency between model outputs ground truth, (2) inherent...
Providing knowledge documents for large language models (LLMs) has emerged as a promising solution to update the static inherent in their parameters. However, document may conflict with memory of LLMs due outdated or incorrect LLMs' This leads necessity examining capability assimilate supplemental external that conflicts memory. While previous studies have explained what extent extract conflicting from provided text, they neglect reason knowledge. Furthermore, there lack detailed analysis on...
Reward models are critical in techniques like Reinforcement Learning from Human Feedback (RLHF) and Inference Scaling Laws, where they guide language model alignment select optimal responses. Despite their importance, existing reward benchmarks often evaluate by asking them to distinguish between responses generated of varying power. However, this approach fails assess on subtle but content changes variations style, resulting a low correlation with policy performance. To end, we introduce...
Information Extraction (IE) aims to extract structural knowledge (e.g., entities, relations, events) from natural language texts, which brings challenges existing methods due task-specific schemas and complex text expressions. Code, as a typical kind of formalized language, is capable describing under various in universal way. On the other hand, Large Language Models (LLMs) trained on both codes texts have demonstrated powerful capabilities transforming into codes, provides feasible solution...
The asymptotic properties of node distribution and speed in random direction model were analyzed, respectively, by the tools geometric probability palm calculus. density function for circular regions was obtained which indicated that mobile nodes tended to disperse as simulation advancing. decay phenomenon confirmed this model. Moreover, hypostasis proved be correlation between duration within any movement period.
Stationary properties of node distribution and speed for random direction model were studied. Firstly, geometric probability was applied to analyze stationary distribution. A closed form density function circular regions obtained which indicated that mobile nodes tended disperse with simulation advancing. Secondly, we palm calculus calculate proved there decay phenomenon in this model. Additionally, a general explanation given such issue.
With this paper, we propose a network coding based cloud storage scheme. The system is in the form of an m * n data array. columns stand for nodes, which are comprised part systematic nodes storing source symbols and nonsystematic parity symbols. Every row array (n, k) Maximum Distance Separable (MDS) code. A symbol only involved encoding with unique row; it locates at not used by other rows. Such design significantly decreases complexity decoding. Moreover, case single node failures, use...
Soldier Radio Waveform (SRW) network is a kind of mobile Ad hoc network. It mostly suitable to tactical or emergency communications. The Time Division Multiple Access (TDMA) control mechanism Barrage Relay Network (BRN) used in SRW. To make simulation study on BRN, group models, node and process models BRN are built with OPNET kits. Moreover, the includes Controlled Region (CBR), which representative application BRN. procedures building CBR data transmission implemented. correctness these...
WSN (Wireless Sensor Network) is constrained by limited radio propagation range as well battery life. Multicast an efficient method to support group communication and decrease communicating energy. It challenging design multicast routing protocol for WSN. The mesh-based ODMRP (On-Demand Routing Protocol) our focus. Its forwarding can be used not only delivering traffic from base station sensor nodes but also reverse the assemble data. feasibility improvement of are analyzed modeling...
In this paper, we firstly study the connectivity of ad hoc networks and establish correlations with average node degree. Then, based on simulation result connectivity, present a probabilistic flooding algorithm that can tradeoff efficiency reliability by using weighting factor. Thus, treat well-known contradictions in our new algorithm. Our results show that, choice an optimal probability factor, get great deal overhead gain little loss packet delivery ratio.