Intracranial Aneurysm (IA) lesion segmentation is significant for IA treatment, which is one of the high death rate and deformity cerebrovascular diseases. Segmenting the IA lesions accurately is still challenging in ...
Intracranial Aneurysm (IA) lesion segmentation is significant for IA treatment, which is one of the high death rate and deformity cerebrovascular diseases. Segmenting the IA lesions accurately is still challenging in digital subtraction angiography (DSA) images due to blurred boundaries, imaging noise, and intracranial vascular morphologies. In this paper, we are the first time to propose a novel instance segmentation network architecture, Graph Mask2Former, to segment IA lesions automatically based on DSA images. Specifically, we apply a graph convolution module to reassign label information, aiming to adjust the confidence weight of error instances adaptively. Furthermore, we design a local refinement module to refine the coarse mask output. The extensive experiments on the clinical IA- DSA and LiTS datasets show that our method outperforms recent state-of-the-art methods. This paper also provides the visual analysis to explain the inherent behavior of our method.
Given the increasingly serious security threats faced by embedded systems, an analysis has been conducted on the characteristics and vulnerabilities of embedded systems. Detailed discussions are provided on security t...
详细信息
Pattern effect as dishing and erosion on Cu-CMP process become a serious problem to fabricate multilevel interconnect in advanced ULSI devices. In the current study, planarization models using patterned wafer and ANSY...
详细信息
In this paper, we address contextual limitations of current deep learning-based and heuristic key phrase extraction tools as applied to the domain of cybersecurity. To address these limitations, we develop a hybrid sy...
In this paper, we address contextual limitations of current deep learning-based and heuristic key phrase extraction tools as applied to the domain of cybersecurity. To address these limitations, we develop a hybrid system that augments state-of-the-art (SOTA) transformers for the task of key phrase sequence labeling, using a novel set of part-of-speech (POS) and role-aware tagging rules to generate fine-grained tag sequences from short text corpora. Next, we fine-tune multiple SOTA deep learning (DL) language model (LM) architectures to these transformed sequences. We then evaluate the architectures by measuring the outcomes from respective LMs to select the best-performing underlying transformers for extracting cybersecurity key phrases. This new ensemble achieves very significant predictive gains over SOTA baselines on general cybersecurity corpora, such as F1 scores at least 25% higher than hybrid SOTA transformers fine-tuned using baseline tagging rules on the generic corpus, with a much less significant tradeoff (of less than 5% in F1) on a vulnerability-specific corpus.
The second harmonic (SH) of an axicon generated Bessel-Gauss beam is created through the nonlinear interaction of photons with a crystal, resulting in the energy doubling of the output photons. In this work, we show e...
详细信息
The second harmonic (SH) of an axicon generated Bessel-Gauss beam is created through the nonlinear interaction of photons with a crystal, resulting in the energy doubling of the output photons. In this work, we show experimentally that in addition to frequency doubling, the SH of Bessel-Gauss beams under asymmetric aberrations from an elliptic axicon exhibit intriguing beam formation. Particularly, the central region of the SH beam profile is composed of two central spots of various geometries surrounded by nested ellipses; one of which is the configuration of two central gamma dots with similar radius knotted by nested ellipses for a zeroth-order Bessel-Gauss pump. These SH beams consistently maintain their spatial profile throughout propagation, reminiscing the behavior of screw dislocations in wave patterns. Our numerical simulations produce beam dynamics consistent with that of experiments and further implicate the remarkable interweaving of bright spots with dark vortices. This is especially noticeable when the beams dynamically oscillate along the optical axis, resulting in the genesis of spatially polarized beams with a knotted framework. The insights gained from our study establish a novel paradigm for exploring interactions of Bessel-like beams with vortex dynamics. This, in turn, has the potential to spark innovations in optical applications, fostering new methodologies to harness and manipulate complex light structures. Our experimental findings could spur new methods of generating logical states of light and new opportunities for material processing control.
Hydraulic systems in production equipment are relatively complex, involving various time series sensor data, such as operating sound frequencies and hydraulic pressure. However, fault prediction generally faces the bo...
Hydraulic systems in production equipment are relatively complex, involving various time series sensor data, such as operating sound frequencies and hydraulic pressure. However, fault prediction generally faces the bottleneck of insufficient fault data, leading to a severe imbalance between normal and fault data ratios. This imbalance makes it difficult for prediction models to effectively learn fault trends. In this paper, we propose a synthesis method for time series data to address this issue. The experimental results demonstrate that the synthesized data lead to significant improvements.
In this paper, a new quasi-resonant DC-DC converter topology is presented, which is the result of combining simpler configurations of single input and single output. The proposed topology is a combination of the Zeta ...
详细信息
This paper addresses planning problems for mobile robots. We consider missions that require accomplishing multiple high-level sub-tasks, expressed in natural language (NL), in a temporal and logical order. To formally...
详细信息
Clustering algorithms are crucial in uncovering hidden patterns and structures within datasets. Among the density-based clustering algorithms, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) has g...
Clustering algorithms are crucial in uncovering hidden patterns and structures within datasets. Among the density-based clustering algorithms, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) has gained considerable attention for its effectiveness in various applications. However, determining appropriate parameter values for this algorithm remains a challenging task. This paper presents a novel methodology for eps parameter estimation for an improved DBSCAN, namely SS-DBSCAN. The experimental results across nine datasets demonstrate the efficacy of our proposed method in accurately determining clusters with eps value from SS-DBSCAN algorithm. The clusters identified using estimated eps values by SS-DBSCAN align well with the inherent structure of the datasets, yielding better cluster results than the manually set parameters and other methods used for automatic estimations of the eps for DBSCAN. Our approach adapted well to the peculiarities of each dataset, whether dealing with different scales, dimensions, or densities; it proved the versatility and robustness across various datasets, thereby emphasizing its generalizability and potential for broader applications.
Multi-Task Learning (MTL) is a learning paradigm that effectively leverages both task-specific and shared information to address multiple related tasks simultaneously. In contrast to Single-Task Learning (STL), MTL of...
详细信息
Multi-Task Learning (MTL) is a learning paradigm that effectively leverages both task-specific and shared information to address multiple related tasks simultaneously. In contrast to Single-Task Learning (STL), MTL offers a suite of benefits that enhance both the training process and the inference efficiency. MTL's key advantages encompass streamlined model architecture, performance enhancement, and cross-domain generalizability. Over the past twenty years, MTL has become widely recognized as a flexible and effective approach in various fields, including computer vision, natural language processing, recommendation systems, disease prognosis and diagnosis, and robotics. This survey provides a comprehensive overview of the evolution of MTL, encompassing the technical aspects of cutting-edge methods from traditional approaches to deep learning and the latest trend of pretrained foundation models. Our survey methodically categorizes MTL techniques into five key areas: regularization, relationship learning, feature propagation, optimization, and pre-training. This categorization not only chronologically outlines the development of MTL but also dives into various specialized strategies within each category. Furthermore, the survey reveals how the MTL evolves from handling a fixed set of tasks to embracing a more flexible approach free from task or modality constraints. It explores the concepts of task-promptable and -agnostic training, along with the capacity for zero-shot learning, which unleashes the untapped potential of this historically coveted learning paradigm. Overall, we hope this survey provides the research community with a comprehensive overview of the advancements in MTL from its inception in 1997 to the present in 2023. We address present challenges and look ahead to future possibilities, shedding light on the opportunities and potential avenues for MTL research in a broad manner. This project is publicly available at https://***/junfish/Awesome-Multi
暂无评论