The primary objective of anomaly detection is to identify abnormal or unusual patterns within a dataset, where the number of normal samples typically exceeds that of abnormal samples. Due to the scarcity of labeled ab...
详细信息
This paper studies and presents data-driven methods for finding rankings of traffic links in a network for optimal traffic data reconstruction based on measurements taken from a subset of links. The link ranking repre...
详细信息
This paper studies and presents data-driven methods for finding rankings of traffic links in a network for optimal traffic data reconstruction based on measurements taken from a subset of links. The link ranking represents the importance of respective links in terms of reconstructing traffic information from sparsely placed sensors, connected vehicles, or other state-of-the-art methods. We first present a baseline method based matrix factorization of the eigen-vector basis matrix, followed by column pivoting. Moreover, we propose a reinforcement learning framework to improve the ranking method when the traffic data is used for the purpose of routing. This study utilizes dynamic traffic data that is observed and estimated from simulation.
Intrinsic and environmental factors contribute to variability in the performance of cells within a battery pack, affecting the lifespan and safety of battery systems. To solve this problem, active and passive equaliza...
详细信息
The aim of this study is to estimate thtrae influence of information and communication technology, based on the use of the ICT development index (measured by the number of Internet users, fixed broadband Internet subs...
The aim of this study is to estimate thtrae influence of information and communication technology, based on the use of the ICT development index (measured by the number of Internet users, fixed broadband Internet subscribers, and the number of mobile subscriptions per 100 inhabitants), on economic development measured by gross domestic product (GDP) per capita. For this purpose, data were collected from a panel of 43 European countries from 2000 to 2020, and three econometric models were used to investigate the impact of ICT use on GDP per capita growth. First, multiple linear regression (MLR) and least absolute shrinkage and selection operator (LASSO) regression were used to research the impact of ICT use on individual countries. Then, a generalized linear dynamic data (GLD) panel model was taken to provide a general model of the dependence of the observed variables. Accordingly, a positive and significant relationship between real GDP per capita and ICT use in Europe is confirmed. The obtained results also show that the impact of ICT on economic growth is greater in low-income countries, i.e., those belonging to Eastern and Central Europe. It is crucial that governments follow the dynamics of the ICT sector and implement specific policies.
A hypergraph H consists of a set V of vertices and a set E of hyperedges that are subsets of V . A t-tuple of H is a subset of t vertices of V . A t-tuple k-coloring of H is a mapping of its t-tuples into k colors. A ...
详细信息
A hypergraph H consists of a set V of vertices and a set E of hyperedges that are subsets of V . A t-tuple of H is a subset of t vertices of V . A t-tuple k-coloring of H is a mapping of its t-tuples into k colors. A coloring is called (t, k, f)-polychromatic if each hyperedge of E that has at least f vertices contains tuples of all the k colors. Let fH(t, k) be the minimum f such that H has a (t, k, f)-polychromatic coloring. For a family of hypergraphs H let fH(t, k) be the maximum fH(t, k) over all hypergraphs H in H. Determining fH(t, k) has been an active research direction in recent years. This is challenging even for t = 1. We present several new results in this direction for t ≥ 2. • Let H be the family of hypergraphs H that is obtained by taking any set P of points in 2, setting V := P and E := {d ∩ P : d is a disk in R2}. We prove that fH(2, k) ≤ 3.7k, that is, the pairs of points (2-tuples) can be k-colored such that any disk containing at least 3.7k points has pairs of all colors. We generalize this result to points and balls in higher dimensions. • For the family H of hypergraphs that are defined by grid vertices and axis-parallel rectangles in the plane, we show that fH(2, k) ≤ √ck ln k for some constant c. We then generalize this to higher dimensions, to other shapes, and to tuples of larger size. • For the family H of shrinkable hypergraphs of VC-dimension at most d we prove that fH(d+1, k) ≤ ck for some constant c = c(d). Towards this bound, we obtain a result of independent interest: Every hypergraph with n vertices and with VC-dimension at most d has a (d+1)-tuple T of depth at least n/c , i.e., any hyperedge that contains T also contains nc other vertices. We also present analogous bounds for coloring pairs of points with respect to pseudo-disks in the plane. • For the relationship between t-tuple coloring and vertex coloring in any hypergraph H we establish the inequality 1/e · tk1/t ≤ fH(t, k) ≤ fH(1, tk1/t ). For the special case of k = 2, ref
We propose a novel optical flow based approach to enhance the axial resolution of anisotropic 3D EM volumes to achieve isotropic 3D reconstruction. Assuming spatial continuity of 3D biological structures in well align...
详细信息
ISBN:
(数字)9798331520526
ISBN:
(纸本)9798331520533
We propose a novel optical flow based approach to enhance the axial resolution of anisotropic 3D EM volumes to achieve isotropic 3D reconstruction. Assuming spatial continuity of 3D biological structures in well aligned EM volumes, we reasoned that optical flow estimation techniques, often applied for temporal resolution enhancement in videos, can be utilized. Pixel-level motion is estimated between neighboring 2D slices along z, using spatial gradient flow estimates to interpolate and generate new 2D slices resulting in isotropic voxels. We leverage recent state-of-the-art learning methods for video frame interpolation and transfer learning techniques, and demonstrate the success of our approach on publicly available ultrastructure EM volumes.
Plant diseases directly affect farm output by lowering grain, fruit, and vegetable quality, threatening global food security. Thus, farmers inspect plant leaves with their eyes. Plant leaf monitoring is unreliable and...
详细信息
ISBN:
(数字)9798331515683
ISBN:
(纸本)9798331515690
Plant diseases directly affect farm output by lowering grain, fruit, and vegetable quality, threatening global food security. Thus, farmers inspect plant leaves with their eyes. Plant leaf monitoring is unreliable and error-prone. A multitude of deep learning algorithms have been developed to detect plant leaf diseases; however, most rely on low-resolution photos utilizing convolutional neural nets (CNNs). This research seeks to create an improved nine-layer the CNN network model for the accurate classification of leaf-related illnesses. A contrast enhancement approach is used to pre-process plant leaf pictures. Following binary thresholding, the pre-processed pictures are separated into leaf images and abnormality segmented using the "Enhanced U-Net (EU-Net)" approach. A Multilevel Feature Fusion Network Convolutional Neural Network (MFFN-CNN) is employed to categorize illnesses of leaves based on segmented images. The "Hybrid Leader Cat Swarm Optimization(HLCSO)" method improves U-Net parameter optimization. The effectiveness of the leaf diagnostic model is demonstrated experimentally employing a variety of factors through the use of numerous baseline approaches.
Vision-language models (VLMs) have made significant progress in reasoning within natural scenes, yet their potential in medical imaging remains largely underexplored. Medical reasoning tasks, which require robust imag...
详细信息
Vision-language models (VLMs) have made significant progress in reasoning within natural scenes, yet their potential in medical imaging remains largely underexplored. Medical reasoning tasks, which require robust image analysis and the generation of well-justified answers, present unique challenges due to the inherent complexity of medical images. Transparency and trustworthiness are critical not only for gaining clinicians’ confidence but also for meeting stringent regulatory requirements. To address these challenges, we propose Med-R1, a novel framework that investigates whether reinforcement learning (RL) can enhance the generalizability and trustworthiness of VLMs in medical reasoning. Building on the recently introduced DeepSeek strategy, we adopt Group Relative Policy Optimization (GRPO) for RL, which encourages models to explore reasoning paths guided by reward signals. Unlike supervised fine-tuning (SFT), which often overfits to training data and struggles with generalization, RL enables models to develop more robust and diverse reasoning capabilities. We comprehensively evaluate Med-R1 across eight distinct medical imaging modalities: Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Ultrasound, Dermoscopy, Fundus Photography, Optical Coherence Tomography (OCT), Microscopy Images, and X-ray Imaging. Compared to the base model, Qwen2-VL-2B, Med-R1 achieves a 29.94% improvement in average accuracy across these modalities and even outperforms Qwen2-VL-72B—a model with 36 times more parameters. To assess model’s generalization abilities, we further test on five different question types: modality recognition, anatomy identification, disease diagnosis, lesion grading, and biological attribute analysis. Med-R1 also demonstrates superior cross-task generalization, outperforming Qwen2-VL-2B by 32.06% and Qwen2-VL-72B in question-type generalization accuracy. These results highlight that RL not only enhances medical reasoning capabilities but also enables p
Process discovery represents a high percentage of the time demanded along the business process management lifecycle. Texts in natural language, or process descriptions, can be explored as a critical information source...
详细信息
Process discovery represents a high percentage of the time demanded along the business process management lifecycle. Texts in natural language, or process descriptions, can be explored as a critical information source. Extracting information automatically from these documents is challenging due to their ambiguity and complexity. While various (semi-)automated solutions are proposed in the literature, functional code access is rare, and tool deployment is almost nonexistent. This research introduces an accessible web interface that allows users to upload process descriptions and interact with the text, automatically identifying the main process elements from the Business Process Model and Notation. Interactions include highlighting elements associated with specific participants. An evaluation experiment with ten potential real-world users was conducted to assess the tool’s benefits. The experiment compared performance in answering questions related to process descriptions across three scenarios: using plain text, using the proposed tool, and using ChatGPT. Results indicated an advantage in using the tool, particularly regarding time consumption. Data collected on ChatGPT usage also provided insights into how users leverage this technology for problem-solving.
暂无评论