检索结果-内蒙古大学图书馆

Local-to-Global Cross-Modal Attention-Aware Fusion for HSI-X Semantic Segmentation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhang, Xuming Yokoya, Naoto Gu, Xingfa Tian, Qingjiu Bruzzone, Lorenzo The International Institute for Earth System Science Nanjing University Nanjing210023 China The Department of Complexity Science and Engineering Graduate School of Frontier Sciences The University of Tokyo Chiba277-8561 Japan The RIKEN Center for Advanced Intelligence Project Tokyo103-0027 Japan The Key Laboratory of Computational Optical Imaging Technology Aerospace Information Research Institute Chinese Academy of Sciences Beijing100094 China The Department of Information Engineering and Computer Science University of Trento Trento38050 Italy

Hyperspectral image (HSI) classification has recently reached its performance bottleneck. Multimodal data fusion is emerging as a promising approach to overcome this bottleneck by providing rich complementary information from the supplementary modality (X-modality). However, achieving comprehensive cross-modal interaction and fusion that can be generalized across different sensing modalities is challenging due to the disparity in imaging sensors, resolution, and content of different modalities. In this study, we propose a Local-to-Global Cross-modal Attention-aware Fusion (LoGoCAF) framework for HSI-X classification that jointly considers efficiency, accuracy, and generalizability. LoGoCAF adopts a pixel-to-pixel two-branch semantic segmentation architecture to learn information from HSI and X modalities. The pipeline of LoGoCAF consists of a local-to-global encoder and a lightweight multilayer perceptron (MLP) decoder. In the encoder, convolutions are used to encode local and high-resolution fine details in shallow layers, while transformers are used to integrate global and low-resolution coarse features in deeper layers. The MLP decoder aggregates information from the encoder for feature fusion and prediction. In particular, two cross-modality modules, the feature enhancement module (FEM) and the feature interaction and fusion module (FIFM), are introduced in each encoder stage. The FEM is used to enhance complementary information by combining the feature from the other modality across direction-aware, position-sensitive, and channel-wise dimensions. With the enhanced features, the FIFM is designed to promote cross-modality information interaction and fusion for the final semantic prediction. Extensive experiments demonstrate that our LoGoCAF achieves superior performance and generalizes well on various multimodal fusion and classification tasks. The code will be made publicly available. Copyright © 2024, The Authors. All rights reserved.

关键词： Remote sensing

LoraMap: Harnessing the Power of LoRA Connections

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Park, Hyeryun Kwak, Jeongwon Jang, Dongsuk Park, Sumin Choi, Jinwook Interdisciplinary Program for Bioengineering Graduate School Seoul National University Korea Republic of Integrated Major in Innovative Medical Science Graduate School Seoul National University Korea Republic of Department of Biomedical Engineering College of Medicine Seoul National University Korea Republic of Medical Research Center Institute of Medical and Biological Engineering Seoul National University Korea Republic of

Fact-checking techniques can mitigate hallucinations in Large Language Models (LLMs), a prominent issue in specialized domains. As parameter-efficient techniques such as Low-Rank Adaptation (LoRA) can overcome substantial computational overhead, some studies have explored the integration of multiple LoRAs. While previous studies focus on parallel integration, this paper investigates methods to establish connections among multiple LoRAs. We create three reasoning datasets tailored to fact-checking and fine-tune individual LoRAs, allowing them to view and reason from diverse perspectives. Then, we explore strategies for allocating these reasoning LoRAs and introduce LoraMap, an approach to map connections between them. The results of the fact-checking task demonstrate that the performance of LoraMap is superior to LoraHub, an existing method for integrating LoRAs. LoraMap also outperforms with significantly fewer trainable parameters than LoraConcat, which concatenates LoRAs and further fine-tunes them. Copyright © 2024, The Authors. All rights reserved.

关键词： Problem oriented languages

Synthetic Batik Pattern Generator Using Wasserstein Generative Adversarial Network with Gradient Penalty

学校读者我要写书评

暂无评论

Synthetic Batik Pattern Generator Using Wasserstein Generati...

International Seminar on Research of Information Technology and Intelligent Systems (ISRITI)

作者： Kus Andriadi Yaya Heryadi Lukas Wayan Suparta Ilvico Sonata Computer Science Department Binus Graduate Program Doctor of Computer Science Bina Nusantara University Jakarta Indonesia Cognitive Engineering Research Group (CERG) Universitas Katolik Indonesia Atma Jaya Jakarta Indonesia Department of Electrical Engineering Faculty of Industrial Technology Institut Teknologi Nasional Yogyakarta Yogyakarta Indonesia

Batik is an Indonesian world cultural heritage. Batik consists of many kinds of patterns depending on where the batik comes from, Batik-making techniques continue to develop along with technology development. Among the batik making techniques that are widely used are hand-written, stamping, and printing. Batik motifs have been widely used as research material, especially in the field of artificial intelligence. The diverse appearance of batik motifs has attracted many researchers to carry out research on making synthetic batik patterns, one of which uses a Generative Adversarial Network. This paper presents a synthetic batik pattern model based on the Wasserstein Generative Adversarial Network with Gradient Penalty. This model has been proven to create new synthetic batik patterns quite well and almost identical with images provided in the dataset, with the notes if the dataset provided is large.

关键词：

Cloud-Native Mobile Application Applied for Law and Regulations Harmonization

学校读者我要写书评

暂无评论

Cloud-Native Mobile Application Applied for Law and Regulati...

Intelligent Cybernetics Technology & Applications (ICICyTA), International Conference on

作者： Muhammad Juan Al Qausar Yaya Heryadi Lukas Lukas Wayan Suparta Ilvico Sonata Computer Science Department Binus Graduate Program Doctor of Computer Science Bina Nusantara University Jakarta Indonesia Cognitive Engineering Research Group (CERG) Universitas Katolik Indonesia Atma Jaya Jakarta Indonesia Department of Electrical Engineering Faculty of Industrial Technology Institut Teknologi Nasional Yogyakarta Yogyakarta Indonesia

Navigating the complex legal and regulatory landscape requires a sophisticated platform that is not only comprehensive but also user-friendly and enables seamless analysis and document comparison in the legal realm. To address this challenging requirement, this research is dedicated to the user experience (UX) design of state-of the-art cloud-native mobile applications specifically designed for legal and regulatory harmonization. The methodology used is based on design thinking principles and takes advantage of the benefits of cloud computing. These include, but are not limited to, scalability, reliability, and security. The result is a solution that is not only powerful but also cost-effective. The core of this application follows a user-centered design philosophy and involves potential users closely in the design process. A comprehensive evaluation is performed using the System Usability Scale (SUS) methodology to measure the effectiveness and usability of the application. The results of this comprehensive approach are convincing. The application has an impressive 95.5% usage rate, indicating strong user engagement. Furthermore, user satisfaction ratings are particularly high and transformative impact of this application in a complex environment of legal harmonization. Through its innovative design, seamless functionality, and user-centric approach, this application has proven to be not only a technical solution, but also a catalyst for positive change in the legal and regulatory harmonization process.

关键词：

Optimized Transmission Strategy for UAV-RIS 2.0 Assisted Communications Using Rate Splitting Multiple Access

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huroon, Aamer Mohamed Huang, Yu-Chih Wang, Li-Chun Department of Electrical Engineering and Computer Science International Graduate Program National Yang Ming Chiao Tung University Taiwan Institute of Communications Engineering National Yang Ming Chiao Tung University Taiwan Department of Electrical and Computer Engineering National Yang Ming Chiao Tung University Taiwan

In this paper, we study the transmission strategy of a ground-based beyond diagonal reconfigurable intelligent surface (BD-RIS), a.k.a RIS 2.0, in a network where multiple unmanned aerial vehicles (UAVs) simultaneously transmit signals to the respective groups of users. It is assumed that each group is assigned subcarriers orthogonal to those assigned to other groups and rate splitting multiple access (RSMA) is adopted within each group. A corresponding mixed integer nonlinear programming problem (MINLP) is formulated, which aims to jointly optimize 1) allocation of BD-RIS elements to groups, 2) BD-RIS phase rotations, 3) rate allocation in RSMA, and 4) precoders. To solve the problem, we propose using generalized benders decomposition (GBD) augmented with a manifold-based algorithm. GBD splits the MINLP problem into two sub-problems, namely the primal and the relaxed master problem, which are solved alternately and iteratively. In the primal problem, we apply block coordinate descent (BCD) to manage the coupling of variables effectively. Moreover, we recognize the manifold structure in the phase rotation constraint of BD-RIS, enabling the Riemannian conjugate gradient (RCG). Simulation results demonstrate the effectiveness of the proposed approach in maximizing spectral efficiency. Copyright © 2023, The Authors. All rights reserved.

关键词： Iterative methods

Densely sampled phylogenies frequently deviate from maximum parsimony in simple and local ways

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Howard-Snyder, William Dumm, Will Barker, Mary Milanov, Ognian Winston, Claris Rich, David H. Matsen, Frederick A. Department of Computer Science and Engineering University of Washington SeattleWA United States Computational Biology Program Fred Hutchinson Cancer Research Center SeattleWA United States Howard Hughes Medical Institute Computational Biology Program Fred Hutchinson Cancer Research Center SeattleWA United States Department of Genome Sciences University of Washington SeattleWA United States Department of Statistics University of Washington SeattleWA United States

Why do phylogenetic algorithms fail when they return incorrect answers? This simple question has not been answered in detail, even for maximum parsimony (MP), the simplest phylogenetic criterion. Understanding MP has recently gained relevance in the regime of extremely dense sampling, where each virus sample commonly differs by zero or one mutation from another previously sampled virus. Although recent research shows that evolutionary histories in this regime are close to being maximally parsimonious, the structure of their deviations from MP is not yet understood. In this paper, we develop algorithms to understand how the correct tree deviates from being MP in the densely sampled case. By applying these algorithms to simulations that realistically mimic the evolution of SARS-CoV-2, we find that simulated trees frequently only deviate from maximally parsimonious trees locally, through simple structures consisting of the same mutation appearing independently on sister branches. Copyright © 2023, The Authors. All rights reserved.

关键词： Phylogenetic

Developments in the utilization of heterogeneous catalysts for biodiesel generation: An update on recent advancements

学校读者我要写书评

暂无评论

Journal of the Taiwan Institute of Chemical Engineers 2024年

作者： Parameshwar, Sai Agnihotri, Mayank Bhan, Uday Nangan, Senthilkumar Verma, Deepak Rajendran, Saravanan Jain, Siddharth Department of Petroleum Engineering and Earth Sciences UPES Dehradun248007 India Metallurgy and Materials Science Research Institute Chulalongkorn University Bangkok10330 Thailand International Graduate Program of Nanoscience and Technology Chulalongkorn University Bangkok10330 Thailand Instituto de Alta Investigacion Universidad de Tarapacá Arica1000000 Chile Department of Mechanical Engineering UPES Dehradun248007 India

Background: Biodiesel is one of the ecofriendly energy sources that can be generated from edible, non-edible, waste, and algae feedstocks. Over time, the utilization of different kinds of heterogeneous catalysts has been proven critical in the manufacture of biodiesel from diverse feedstocks. They exhibit better reusability and easier separation compared to the homogenous catalysts. Methods: In this work, the advances in the use of heterogeneous catalysts for biodiesel production from 2018 to 2024 have been evaluated. The process parameters like alcohol to oil molar ratio (M/O), catalyst concentration (%), reaction time and reaction temperature (RT), reusability of the catalyst, etc. are evaluated. The catalysts derived from waste materials through the calcination process are an efficient and cost-effective substitute for conventional catalysts. Nanomaterials, due to their small size and higher aspect ratio, provide very high efficiency and reusability in catalytic reactions. Significant findings: The heterogeneous catalysts are also found to be highly reusable for multiple cycles and the leaching of active sites, clogging of pore spaces of catalyst, etc. are the major causes of catalyst deactivation. Magnetic heterogeneous catalysts have shown good biodiesel yields and high reusability. Also, naturally occurring calcium oxide and potassium compounds present in waste derived catalysts helped to give high biodiesel yields. The future direction in the path of development of heterogeneous catalysts from cheap, abundant, and environmentally benign materials. The synthesis of heterogeneous catalysts from waste materials addresses the issue of waste disposal and converting solid waste to value-added catalysts. The life cycle assessment studies, encompassing heterogeneous catalyst synthesis, feedstock oil extraction, biodiesel production, and fuel combustion may give valuable input on the environmental impact of the entire process. © 2024

关键词： Potassium compounds

Thermodynamics of the Solid-Liquid Phase Equilibrium of a Binary System: Effect of a Chemical Reaction in the Liquid and Epitaxial Strain in the Solid

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Deymier, Alix C. Deymier, Pierre Alix Muralidharan, Krishna Latypov, Marat I. Department of Biomedical Engineering UConn Health FarmingtonCT06030 United States Department of Materials Science and Engineering University of Arizona TucsonAZ85721 United States Graduate Interdisciplinary Program in Applied Mathematics University of Arizona TucsonAZ85721 United States

We introduce two thermodynamics models of solid-liquid phase equilibrium of a binary system in the presence of (a) a chemical reaction in the liquid involving only one of the binary components and, (b) strain-inducing epitaxial conditions on the solid in addition to the chemical reaction in the liquid. These models show the existence of the counter intuitive behavior that epitaxy may promote recrystallization of the solid in spite of the reaction driving dissolution. The conditions for this recrystallization behavior are determined by the value of the epitaxial lattice matching composition, the temperature and the overall composition of the binary solid-liquid system. © 2023, The Authors. All rights reserved.

关键词： Chemical reactions

Generalization is not a universal guarantee: Estimating similarity to training data with an ensemble out-of-distribution metric

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Schreyer, W. Max Anderson, Christopher Thompson, Reid F. Computational Biology Program Oregon Health & Science University PortlandOR United States Department of Biomedical Engineering Oregon Health & Science University PortlandOR United States Portland VA Research Foundation PortlandOR United States Department of Radiation Medicine Oregon Health & Science University PortlandOR United States Department of Medical Informatics and Clinical Epidemiology Oregon Health & Science University PortlandOR United States Division of Hospital and Specialty Medicine VA Portland Healthcare System PortlandOR United States

Failure of machine learning models to generalize to new data is a core problem limiting the reliability of AI systems, partly due to the lack of simple and robust methods for comparing new data to the original training dataset. We propose a standardized approach for assessing data similarity in a model-agnostic manner by constructing a supervised autoencoder for generalizability estimation (SAGE). We compare points in a low-dimensional embedded latent space, defining empirical probability measures for k-Nearest Neighbors (kNN) distance, reconstruction of inputs and task-based performance. As proof of concept for classification tasks, we use MNIST and CIFAR-10 to demonstrate how an ensemble output probability score can separate deformed images from a mixture of typical test examples, and how this SAGE score is robust to transformations of increasing severity. As further proof of concept, we extend this approach to a regression task using non-imaging data (UCI Abalone). In all cases, we show that out-of-the-box model performance increases after SAGE score filtering, even when applied to data from the model’s own training and test datasets. Our out-of-distribution scoring method can be introduced during several steps of model construction and assessment, leading to future improvements in responsible deep learning implementation. Copyright © 2025, The Authors. All rights reserved.

关键词：