In the context of industrial upgrading, the exponential growth of computational demands for big data processing has emerged as a critical challenge. Cloud computing, with its evolving technological capabilities, prese...
详细信息
ISBN:
(数字)9798331536169
ISBN:
(纸本)9798331536176
In the context of industrial upgrading, the exponential growth of computational demands for big data processing has emerged as a critical challenge. Cloud computing, with its evolving technological capabilities, presents a viable solution to address these escalating computational requirements. This study focuses on optimizing resource allocation efficiency through the establishment of a multiobjective optimization framework for cloud computing systems. Specifically, we formulate mathematical representations of objective functions and constraints while detailing corresponding encoding/decoding methodologies. The research employs the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to resolve task scheduling dilemmas in cloud environments, simultaneously optimizing makespan and total computational time through innovative applications of nondominated sorting and congestion distance comparison. A series of multi-scale comparative experiments demonstrate the algorithm's superior performance over conventional scheduling approaches.
Generative Adversarial Networks (GAN) involve the competition between a generator and a discriminator. The large variance of training data can bring difficulty to GAN resulting in mode collapse, training instability a...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Generative Adversarial Networks (GAN) involve the competition between a generator and a discriminator. The large variance of training data can bring difficulty to GAN resulting in mode collapse, training instability and low quality. As the distribution of generated samples evolves, the discriminator will have different levels of confidence for its predictions. To mitigate these problems, we introduce confidence-aware discrimination (CAD) to guide the training and sampling of GANs. To adapt confidence estimation for GANs, we design a new architecture based on StyleGAN2, and propose a confidence-aware adversarial loss. Extensive experiments are conducted on face, scene and object generation benchmarks. In the training stage, CAD can progressively learn the training data with the guide of confidence estimation and lead to a better convergence. In the inference stage, the confidence score can provide a new dimension of metric to assess the quality of generated samples and can further improve the performance by resampling and finetuning.
Convolutional sparse representation (CSR) extends the standard form by incorporating convolutional operations and has achieved notable success in image processing applications, offering a shift-invariant representatio...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
Convolutional sparse representation (CSR) extends the standard form by incorporating convolutional operations and has achieved notable success in image processing applications, offering a shift-invariant representation model. Recent studies typically redesign the initial model by exploiting the convolution-multiplication property of the discrete Fourier transform to solve the involved convex optimizations efficiently, which imposes periodic boundary handling and risks introducing boundary artifacts. This paper generalizes this approach by leveraging 46 × 46 convolution-multiplication properties while requiring only extra element-wise operations, enabling model designs that combine periodic, antiperiodic, and 40 symmetric boundary handlings for each dimension and across convolutions. These newly accepted models demonstrated better representation performance than the conventional model in our experiments, indicating the benefits of selecting appropriate boundary handlings.
Document categorization continues to be a significant area of research, particularly in the context of automating the indexing of diverse web content such as blogs and forums. However, large document classification po...
详细信息
ISBN:
(数字)9798331542726
ISBN:
(纸本)9798331542733
Document categorization continues to be a significant area of research, particularly in the context of automating the indexing of diverse web content such as blogs and forums. However, large document classification poses challenges in relation with both performance and processing time. This paper empirically investigates the impact of text summarization on document classification. For that, we compare classification performance before and after applying summarization techniques, focusing on two key aspects: (1) computational time and (2) classification performance, measured through accuracy, loss, and F1-score metrics. Our findings demonstrate the effectiveness of the proposed method, highlighting its potential to enhance the efficiency and accuracy of document categorization processes.
An important aspect of deploying face recognition (FR) algorithms in real-world applications is their ability to learn new face identities from a continuous data stream. However, the online training of existing deep n...
详细信息
ISBN:
(数字)9798331510831
ISBN:
(纸本)9798331510848
An important aspect of deploying face recognition (FR) algorithms in real-world applications is their ability to learn new face identities from a continuous data stream. However, the online training of existing deep neural network-based FR algorithms, which are pre-trained offline on large-scale stationary datasets, encounter two major challenges: (I) catastrophic forgetting of previously learned identities, and (II) the need to store past data for complete retraining from scratch, leading to significant storage constraints and privacy concerns. In this paper, we introduce CLFace, a continual learning framework designed to preserve and incrementally extend the learned knowledge. CLFace eliminates the classification layer, resulting in a resource-efficient FR model that remains fixed throughout lifelong learning and provides label-free supervision to a student model, making it suitable for open-set face recog-nition during incremental steps. We introduce an objective function that employs feature-level distillation to reduce drift between feature maps of the student and teacher models across multiple stages. Additionally, it incorpo-rates a geometry-preserving distillation scheme to maintain the orientation of the teacher model's feature embedding. Furthermore, a contrastive knowledge distillation is incor-porated to continually enhance the discriminative power of the feature representation by matching similarities between new identities. Experiments on several benchmark FR datasets demonstrate that CLFace outperforms baseline approaches and state-of-the-art methods on unseen identities using both in-domain and out-of-domain datasets.
In this paper, we present a Situational Awareness based Resource Allocation (SA-RA) strategy for multi-target tracking (MTT) in distributed radar networks, where both the resource utilization and overall MTT accuracy ...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
In this paper, we present a Situational Awareness based Resource Allocation (SA-RA) strategy for multi-target tracking (MTT) in distributed radar networks, where both the resource utilization and overall MTT accuracy can be improved. The fusion rule with probabilistic data association (PDA) is used to associate cumulative data with measurement data from radar nodes. We also further derive the Bayesian Cramér-Rao Lower Bound (BCRLB) under the PDA fusion rule. For the proposed SA-RA strategy, the targets behavior is analyzed to determine their importance weights, expected tracking accuracy, and the allocated beams for each target. Utilizing the PDA-BCRLB and the importance weights, the objective function of the SA-RA strategy is formulated as a weighted sum of target utility functions. By addressing this objective function, the optimal transmission power is determined. Simulation results verify the superiority both in terms of tracking performance and resource allocation.
This paper presents a new graph-cuts based volumetric reconstruction technique for multi-view stereo. To refrain from the balloon constraint term widely used in 3D reconstruction approaches, we propose the using of a ...
详细信息
This paper presents a new graph-cuts based volumetric reconstruction technique for multi-view stereo. To refrain from the balloon constraint term widely used in 3D reconstruction approaches, we propose the using of a more sophisticated and high performance map, namely the distance fidelity map, for graph-cuts optimization. The approach first seeks to compute the optimal Distance fidelity map (DFM) of each view using photo-consistency, feature constraint, silhouette constraint and smoothness constraint. Occlusion robust and feature robust mechanism are also considered during this process. After this, according to our proposed objective function, multiple DFMs are fused into a graph-cuts framework to give a watertight surface that best capture the surface detail. Results demonstrate the high performance reconstruction quality using only 10 views and its generality to extensive actor structures.
The purpose of this article is to introduce two indicators designed to evaluate and consequently influence the indirect emissions related to the operation of an EV. The first indicator, denoted as “Electric Vehicle C...
详细信息
The purpose of this article is to introduce two indicators designed to evaluate and consequently influence the indirect emissions related to the operation of an EV. The first indicator, denoted as “Electric Vehicle CarbonFlex Potential” indicator, evaluates an EV's maximum and minimum achievable carbon emissions using an optimization approach and compares the user's resultant indirect carbon emissions to these boundaries, therefore, this indicator compares the users behavior to the optimal best and worst cases. The second indicator is “EcoCharge Time” indicator, which provides feedback to an EV user based on their charging behavior on the best and worst times of charging the vehicle in a day. Since human behavior cannot be controlled, such indicators are essential tools for influencing the behavior of EV users toward a desired optimal, in this case, a charging schedule with the lowest possible overall indirect emissions. The proposed indicators were tested on an EV dataset using the carbon intensity data from a number of countries and the results show that there exists considerable flexibility potential. Additionally, the results also showed the best charging times, which were typically clustered around, allowing for ease of use and understanding.
Over the last few decades, many researchers opt various techniques to achieve optimum time for coordination of overcurrent relays. The main motive of these techniques is to avoid mal operation among the main and secon...
详细信息
ISBN:
(数字)9798331520540
ISBN:
(纸本)9798331520557
Over the last few decades, many researchers opt various techniques to achieve optimum time for coordination of overcurrent relays. The main motive of these techniques is to avoid mal operation among the main and secondary relays and to reduce time of operation of each relay as per required criteria. Some researchers have suggested modification in relay characteristics and objective functions to avoid mis-coordination. Three objective functions from the literature are examined in this paper, and new modified objective functions are the suggested after modification. A comparison is made between the standard objective function with modified obje ctive functions. Out of three modified objective functions, the best modified objective function is proposed. Further, three IEC standard relay characteristics have been adopted to analyze & compare the effectiveness & robustness of said objective functions. For the said comparison IEEE 14-bus modified distribution system is considered and a Harmony search algorithm is used.
Enhancing back-drivability in robot joints is crucial for safe and effective physical human-robot interaction (pHRI). This paper presents a novel approach to improve back-drivability by reducing reflected inertia thro...
详细信息
ISBN:
(数字)9798331531614
ISBN:
(纸本)9798331531621
Enhancing back-drivability in robot joints is crucial for safe and effective physical human-robot interaction (pHRI). This paper presents a novel approach to improve back-drivability by reducing reflected inertia through a multi-motor system (MMS). Unlike traditional high-ratio gearboxes, which amplify motor inertia and reduce efficiency, the MMS distributes torque across multiple motors, allowing for lower gear ratios and less reflected inertia. We developed an optimization model considering motor selection, gear ratios, and gear stages to minimize reflected inertia while meeting load and geometric constraints. A case study on an industrial robot’s shoulder joint demonstrates an 88.84% reduction in reflected inertia using the proposed MMS compared to a conventional single-motor system. The findings suggest that multi-motor systems can significantly reduce reflected inertia, improving back-drivability in robot joints for safer and more efficient pHRI.
暂无评论