the proceedings contain 23 papers. the special focus in this conference is on Skin Imaging Collaboration, Interpretability of Machine Intelligence in Medical Image computing, Embodied AI and Robotics for Healthcare Wo...
ISBN:
(纸本)9783031776090
the proceedings contain 23 papers. the special focus in this conference is on Skin Imaging Collaboration, Interpretability of Machine Intelligence in Medical Image computing, Embodied AI and Robotics for Healthcare Workshop and MICCAI Workshop on distributed, Collaborative and Federated Learning. the topics include: DeCaF 2024 Preface;i2M2Net: Inter/Intra-modal Feature Masking Self-distillation for Incomplete Multimodal Skin Lesion Diagnosis;from Majority to Minority: A Diffusion-Based Augmentation for Underrepresented Groups in Skin Lesion Analysis;segmentation Style Discovery: Application to Skin Lesion Images;a Vision Transformer with Adaptive Cross-Image and Cross-Resolution Attention;lesion Elevation Prediction from Skin Images Improves Diagnosis;DWARF: Disease-Weighted Network for Attention Map Refinement;PIPNet3D: Interpretable Detection of Alzheimer in MRI Scans;Detecting Unforeseen Data Properties with Diffusion Autoencoder Embeddings Using Spine MRI Data;interpretability of Uncertainty: Exploring Cortical Lesion Segmentation in Multiple Sclerosis;TextCAVs: Debugging Vision Models Using Text;evaluating Visual Explanations of Attention Maps for Transformer-Based Medical Imaging;Exploiting XAI Maps to Improve MS Lesion Segmentation and Detection in MRI;EndoGS: Deformable Endoscopic Tissues Reconstruction with Gaussian Splatting;VISAGE: Video Synthesis Using Action Graphs for Surgery;a Review of 3D Reconstruction Techniques for Deformable Tissues in Robotic Surgery;SurgTrack: CAD-Free 3D Tracking of Real-World Surgical Instruments;MUTUAL: Towards Holistic Sensing and Inference in the Operating Room;Complex-Valued Federated Learning with Differential Privacy and MRI Applications;enhancing Privacy in Federated Learning: Secure Aggregation for Real-World Healthcare Applications;federated Impression for Learning withdistributed Heterogeneous Data;A Federated Learning-Friendly Approach for Parameter-Efficient Fine-Tuning of SAM in 3D Segmentation;probing the Effic
the proceedings contain 7 papers. the topics discussed include: containerized bioinformatics ecosystem for HPC;pyp2pcluster: a cluster discovery tool;analysis of user-support tickets in the lifetime of the blue waters...
ISBN:
(纸本)9781665463492
the proceedings contain 7 papers. the topics discussed include: containerized bioinformatics ecosystem for HPC;pyp2pcluster: a cluster discovery tool;analysis of user-support tickets in the lifetime of the blue waters system;interactive NLU-powered ontology-based workflow synthesis for FAIR support of HPC;NERSC job script generator;PMT: power measurement toolkit;and CloudQ: a secure AI / HPC cloud bursting system.
the proceedings contain 6 papers. the topics discussed include: analysis of validating and verifying OpenACC compilers 3.0 and above;OmpSs-2 and OpenACC interoperation;extending MAGMA portability with OneAPI;KokkACC: ...
ISBN:
(纸本)9781665490191
the proceedings contain 6 papers. the topics discussed include: analysis of validating and verifying OpenACC compilers 3.0 and above;OmpSs-2 and OpenACC interoperation;extending MAGMA portability with OneAPI;KokkACC: enhancing Kokkos with OpenACC;SPEL: software tool for porting E3SM land model with OpenACC in a function unit test framework;and GPU-accelerated sparse matrix vector product based on element-by-element method for unstructured FEM using OpenACC.
the proceedings contain 114 papers. the topics discussed include: accounting and monitoring infrastructure for distributedcomputing in the atlas experiment;the atlas EVENTINDEX using the HBase/phoenix storage solutio...
the proceedings contain 114 papers. the topics discussed include: accounting and monitoring infrastructure for distributedcomputing in the atlas experiment;the atlas EVENTINDEX using the HBase/phoenix storage solution;offline software and computing for the SPD experiment;the grid-characteristic method for applied dynamic problems of fractured and anisotropic media;fractal thermodynamics, big data and its 3D visualization;the technology and tools for the building of information exchange package based on semantic domain model;participation of Russian institutes in the processing and storage of ALICE data;a virtual testbed for optimizing the performance of a new type of accelerators;multithreaded event simulation in the BMNROOT package;and resource management in private multi-service cloud environments.
Withthe rapid increase of data, the scale of cloud is gradually expanding, forming a wide-area cloud platform consisting of multiple data centers distributed across different locations. For industry professionals, ma...
详细信息
CFN(Compute First networking) is an intelligent new network with computation at its core and networking as its foundation, integrating Computational, storage, and transmission resources. Computational resources and ne...
详细信息
ISBN:
(纸本)9798350350227;9798350350210
CFN(Compute First networking) is an intelligent new network with computation at its core and networking as its foundation, integrating Computational, storage, and transmission resources. Computational resources and network resources in the CFN are ubiquitous and heterogeneous. In the traditional Internet identifier system, IP addresses have limited extension, poor semantic scalability, making it challenging to incorporate information about computational and network resources into the network and hindering unified description. this paper introduces a multi-dimensional identifier system-based mechanism for computational and network resource awareness. It perceives computational network resources and employs service identifier for unified mapping. Additionally, it designs workflows for computational and network resource awareness, accomplishing tasks such as computational resource notification, request, query, update, as well as network resource detection, collection, and feedback. this facilitates functionality for service localization and computational network resource awareness. Finally, a prototype system was constructed based on the programmable data plane to experimentally evaluate the functionality and performance of the resource awareness mechanism. the results indicate that in a CFN scenario, this mechanism can support real-time perception and updates of computational network resources. In the context of distributedcomputing services, compared to traditional network deployments, the mechanism reduces latency by approximately 10%.
Locally Repairable codes are used in the distributed storage system to minimize the I/O overhead. In this paper, we design topology-aware locally repairable codes based on the network topology of distributed storage s...
详细信息
the surge in ubiquitous computing resources has propelled the evolution of computing-aware networking. Nonetheless, the majority of existing computing service discovery and scheduling mechanisms suffer from deficienci...
详细信息
ISBN:
(纸本)9798350350227;9798350350210
the surge in ubiquitous computing resources has propelled the evolution of computing-aware networking. Nonetheless, the majority of existing computing service discovery and scheduling mechanisms suffer from deficiencies such as suboptimal search efficiency and algorithmic shortcomings, leading to low solution quality and efficiency. In response, this paper proposes a computing service discovery and scheduling mechanism founded on multi-dimensional unified identifier. Initially, a computing service architecture rooted in multi-dimensional unified identifier is outlined. this architecture delineates computing service attributes using multi-dimensional descriptors, enabling service addressing predicated on mapped service identifier. Subsequently, the matching discovery mechanism for computing services is structured around the K-Dimensional Tree (K-D Tree) framework. Moreover, a computing service scheduling mechanism is devised, catering to both single-node and combined service scheduling scenarios. Finally, the experimental test evaluation results show that the discovery delay associated withcomputing services satisfies the requirements across various service request scenarios. Moreover, when contrasted withthe nearby greedy strategy, the proposed computing service scheduling mechanism achieves a superior user satisfaction score.
High speed internet and advanced networking technology contribute to having large number of various edge devices in heterogeneous edge-cloud systems. In conventional cloud computing systems, all device data is process...
详细信息
ISBN:
(纸本)9798350366495;9798350366488
High speed internet and advanced networking technology contribute to having large number of various edge devices in heterogeneous edge-cloud systems. In conventional cloud computing systems, all device data is processed in the centralized cloud servers. the growing number of devices, i.e., increasing amount of device data, poses a challenge to the cloud servers to process data in a time- and energy-efficient manner. Studies show promise to reduce execution time and energy consumption by introducing collaborative edge-cloud computing paradigm. In this work, we study collaborative edge-cloud computing by introducing a framework of pairing the computations at edge and cloud resources to minimize execution time and energy consumption. First, the cloud servers (CSs) are made about 90% utilized by adjusting the device data i. e., computed data. then, each edge server (ES) is optimized using 50% or less of the previously generated device data i.e., cloud computed data. Finally, computations (i.e., device data) are distributed among the ESs and CSs, and performance is assessed to obtain the optimal pairing of computations. A heterogeneous system with one CS, two ESs, 10 edges, and 30 devices of five different types is modeled and simulated using VisualSim. Experimental results show that the proposed method helps reduce execution time and energy consumption by 90% and 56%, respectively. the proposed framework holds a promise for enhancing the scalability of heterogeneous systems, an avenue we intend to explore in our upcoming venture.
Deep learning has become promising across numerous fields in transforming conventional paradigms into smart eras in distributed applications. Large neural networks in recent years have been popular in solving massive ...
详细信息
ISBN:
(纸本)9798331511425;9798331511432
Deep learning has become promising across numerous fields in transforming conventional paradigms into smart eras in distributed applications. Large neural networks in recent years have been popular in solving massive real-world problems. However, the challenge behind the increasing complexity of deep neural networks impacts the training time. Appropriate resource provisioning and rightsizing is the requirement in all standard platforms like the cloud to handle this performance degradation. this research explores distributed CPU clusters as a scalable and cost-effective alternative for training large neural networks. the experiments on two different multi-processing machines with workers' distributions demonstrated the change in maximum accuracies is in a range of 92.96% to 96.74%. As our approach can be adopted and experiments can be extended to serverful and serverless computing training workloads, deep learning researchers and practitioners will benefit from our solution.
暂无评论