Recently, Docker Container as a Service (CaaS) has been provided for multi-user services in the 5G Multi-Access Edge computing (MEC) environment, and servers that support accelerators such as GPUs, not conventional CP...
详细信息
ISBN:
(纸本)9783031503849;9783031503856
Recently, Docker Container as a Service (CaaS) has been provided for multi-user services in the 5G Multi-Access Edge computing (MEC) environment, and servers that support accelerators such as GPUs, not conventional CPU servers, are being considered. In addition, as the number of AI services is increasing and the computation power required by deep neural network model increases, offloading to edge servers is required due to insufficient computational capacity and heat problem of user devices (UE). However, there is a resource scheduling problem because all users' packets cannot be offloaded to the edge server due to resource limitations. To address this problem, we suggest deep reinforcement learning-based GPU CaaS Packet scheduling named as Delcas for stabilizing quality of AI experience. First, we design the architecture using containerized target AI application on MEC GPUs and multiple users send video stream to MEC server. We evaluate video stream to identify the dynamic amount of resource requirement among each users using optical flow and adjust user task queue. To satisfy equal latency quality of experience, we apply lower quality first serve approach and respond hand pose estimation results to each user. Finally, we evaluate our approach and compare to conventional scheduling method in the aspect of both accuracy and latency quality.
This article designs a classification method for long educational news texts based on labeled educational news reports. Firstly, CNN is used to extract local features of text information. Then, a bidirectional long sh...
详细信息
Agriculture is the most significant industry in the economy of India. Various kinds of diseases affect the leaves of plants and influence the productivity of crops. Apple farmers are also constantly facing challenges ...
详细信息
The proceedings contain 192 papers. The topics discussed include: application of artificial intelligence and big data in smart buildings;engine detection and online monitoring technology based on image recognition;mec...
ISBN:
(纸本)9798350395631
The proceedings contain 192 papers. The topics discussed include: application of artificial intelligence and big data in smart buildings;engine detection and online monitoring technology based on image recognition;mechanical parts life prediction and health monitoring system based on deep learning;research on optimal zoning of energy Internet source load matching power grid based on improved K-means algorithm;design of power load forecasting algorithm for distribution network based on machine learning;analysis of sea ice area fluctuation in the arctic circle based on big data and SARIMA model;a redesign method for embroidery pattern graphics based on multiscale image processing technology;and research on data mining of university management decision support archives based on cloud computing.
This paper presents a novel technique for enhancing the allocation of resources in the charging infrastructure for e-bikes by employing deep reinforcement learning (DRL) in a context-specific manner. With the evolving...
详细信息
Our project pioneers a transformative platform, marrying artificial intelligence with fashion customization. Users engage with an intuitive interface, seamlessly translating their style visions into unique apparel des...
详细信息
A major global health concern is Melanoma, a Malignant form of skin cancer. This type of skin cancer is the most aggressive and has the highest fatality rate. An early diagnosis and course of treatment is crucial. Wit...
详细信息
Despite some application of Contextual teaching and learning (CTL) theories in practice, their effectiveness has not been prominent, necessitating comprehensive optimization. To enhance the quality and the success of ...
详细信息
We present a novel reservoir computing approach in which single nonlinear device - semiconductor optical amplifier, replaces the entire nonlinear reservoir to perform computations, without the conventional use of a de...
详细信息
Recently, major advancements have been achieved in the object detection algorithms for behavior detection, but detection for student behavior in low-resolution scenarios remains a great challenge. Aiming at the task o...
详细信息
暂无评论