the Traffic Management System employs Braess' paradox to reduce congestion by altering driving patterns and utilizes the Frank-Wolfe method to create efficient routes. Integrated with a state-of-the-art pothole de...
ISBN:
(纸本)9798400709227
the Traffic Management System employs Braess' paradox to reduce congestion by altering driving patterns and utilizes the Frank-Wolfe method to create efficient routes. Integrated with a state-of-the-art pothole detection model, which utilizes Vision Transformer and Transformer architectures, the system dynamically adapts to road conditions. this innovative approach, trained on a Mumbai Hyperloop station dataset, includes a Two-Way Transformer network for precise road condition analysis, enhancing traffic flow and safety. It has achieved a 35% decrease in traffic congestion, 47% faster incident response, and 33% shorter trip durations. the system's integration of advanced technologies paves the way for future autonomous vehicle networks and offers significant benefits for urban transportation and planning, simultaneously addressing the challenges of potholes and congestion.
An adaptive narrowband beamforming method based on proportional-integral-differential (PID) virtual interference control which is called PID-optimized Olen Campton beamforming method (PIDBF) is proposed, achieving hig...
详细信息
ISBN:
(数字)9798350355895
ISBN:
(纸本)9798350355901
An adaptive narrowband beamforming method based on proportional-integral-differential (PID) virtual interference control which is called PID-optimized Olen Campton beamforming method (PIDBF) is proposed, achieving high-precision and rapid low sidelobe beamforming. In the proposed algorithm, PID controlling gains are applied to iteratively adjust the virtual noise sources. thus, the Olen Campton beamforming method (OBF) is optimized by innovatively introducing PID control method in the iterative process of virtual interference sources, leading to faster and more stable sidelobe control. the superior performance of the proposed method is verified through a number of simulations.
Multi-frame video super-resolution(VSR) aims to restore a high-resolution video from both its corresponding low-resolution frame and multiple neighboring frames, in order to make full use of the inter-frame informatio...
详细信息
ISBN:
(数字)9781665482233
ISBN:
(纸本)9781665482233
Multi-frame video super-resolution(VSR) aims to restore a high-resolution video from both its corresponding low-resolution frame and multiple neighboring frames, in order to make full use of the inter-frame information. However, vast computation complexity hinders the inference speed of video super-resolution. In order to increase the inference speed while ensuring the accuracy of the model, we proposed an efficient and parallel multi-frame VSR network, termed EPVSR. the proposed EPVSR is based on spatio-temporal adversarial learning to achieve temporal consistency and uses TecoGAN as the baseline model. By adding an improved non-deep network, which is composed of parallel subnetworks with multi-resolution streams, these streams are fused together at regular intervals to exchange information. we reduced the number of parameters and make the model lighter. Besides, we implement structural re-parameterization network acceleration technique to optimize the inference process of EPVSR network. Finally, our EPVSR achieves the real-time processing capacity of 4K@36.45FPS. compared with TecoGAN, we achieve 9.75 x performance speedups, but the effect is not reduced. the PSNR of EGVSR are increased by 3.36%. the experimental results show that the nondeep network can effectively speed up the model inference, and the proposed EPVSR has a good super-resolution effect.
Given the swift advancements in GPU architectures, the process of assessing and improving GPU rendering performance has grown increasingly sophisticated and vital. To address these challenges and provide microarchitec...
详细信息
ISBN:
(数字)9798331522216
ISBN:
(纸本)9798331522223
Given the swift advancements in GPU architectures, the process of assessing and improving GPU rendering performance has grown increasingly sophisticated and vital. To address these challenges and provide microarchitecture-agnostic insights, this paper introduces MICPAT, a tool for GPU characteristic profiling. MICPAT extracts key program characteristics such as instruction composition, basic block count, instruction frequency and memory allocation size across NVIDIA's Kepler, Maxwell, Pascal, and Volta GPU series. By analyzing these microarchitecture-agnostic characteristics, developers gain deep insights into the behavior and performance of their GPU programs. MICPAT supports precompiled applications utilizing CUDA, OpenACC, OpenCL, or CUDA Fortran. Serving as a versatile platform, MICPAT enables consistent analysis across this diverse set of GPU architectures and precompiled application environments. Utilizing Octane renderer, as well as Rodinia and Parboil benchmarks, extensive experimental evaluations across 100 GPU rendering applications have validated MICPAT's efficacy and its microarchitecture-agnostic nature. the open source repo is https://***/records/13623324.
the pervasiveness of information and communications technologies, such as 5 G and edge AI, has led to the booming growth of video applications, particularly in the realm of virtual reality (VR). However, due to the ra...
详细信息
ISBN:
(数字)9798350363951
ISBN:
(纸本)9798350363968
the pervasiveness of information and communications technologies, such as 5 G and edge AI, has led to the booming growth of video applications, particularly in the realm of virtual reality (VR). However, due to the random distribution of the video traffic and the fluctuation of the computing power in Mobile Edge Computing (MEC) networks, traditional load balancing strategies lead to high video service response delay and low resource utilization. To this end, this paper designs and implements an adaptive dynamic load-balancing strategy for VR. Based on the performance monitoring function of the Kubernetes (k8s) system, we first utilize the reinforcement learning algorithm TD3 to find the optimal video service allocation at different edge nodes. then, considering the real-time GPU utilization of each edge computing node, the queue length, and the average transmission delay, we jointly optimize the average response delay of users and the work efficiency of nodes. Finally, we evaluate the proposed scheme on a real k8s system. Compared with traditional load balancing strategies, the proposed scheme reduces the latency by $16 \%$ to $52 \%$ and improves the work efficiency by $5 \%$ to $8 \%$.
the widespread adoption of deep neural networks in machine learning calls for an objective quantification of esoteric trust. In this paper we propose GradTrust, a classification trust measure for large-scale neural ne...
详细信息
ISBN:
(数字)9798350351422
ISBN:
(纸本)9798350351439
the widespread adoption of deep neural networks in machine learning calls for an objective quantification of esoteric trust. In this paper we propose GradTrust, a classification trust measure for large-scale neural networks at inference. the proposed method utilizes variance of counterfactual gradients, i.e. the required changes in the network parameters if the label were different. We show that GradTrust is superior to existing techniques for detecting misprediction rates on 50000 images from ImageNet validation dataset. Depending on the network, GradTrust detects images where either the ground truth is incorrect or ambiguous, or the classes are co-occurring. We extend GradTrust to Video Action Recognition on Kinetics-400 dataset. We showcase results on 14 architectures pretrained on ImageNet and 5 architectures pretrained on Kinetics-400. We observe the following: (i) simple methodologies like negative log likelihood and margin classifiers outperform state-of-the-art uncertainty and out-of-distribution detection techniques for misprediction rates, and (ii) the proposed GradTrust is in the Top-2 performing methods on 37 of the considered 38 experimental modalities. the code is available at: https://***/olivesgatech/GradTrust
Underwater target recognition is one of the most challenging tasks in underwater signal processing. Previous deep learning methods have relied on fusing more acoustic features, ignoring the rich information contained ...
Underwater target recognition is one of the most challenging tasks in underwater signal processing. Previous deep learning methods have relied on fusing more acoustic features, ignoring the rich information contained in the time-frequency features of underwater acoustics. Furthermore, fusing features that are less relevant to the target may result in redundancy and affect the recognition performance of the model. In this paper, a novel method based on Multi-Scale feature extraction, Attention mechanism for feature fusion and Convolutional Recurrent Neural Network (MACRN) is proposed to fully exploit the deep features of Mel-Frequency Cepstral Coefficients (MFCC) for underwater target recognition. In comparative experiments conducted on the open-source dataset ShipsEar, the proposed method achieves an average recognition accuracy of 98.1%
the proceedings contain 24 papers. the special focus in this conference is on parallel and Distributed processing Techniques. the topics include: parallel N-Body Performance Comparison: Julia, Rust, and More;REFT...
ISBN:
(纸本)9783031856372
the proceedings contain 24 papers. the special focus in this conference is on parallel and Distributed processing Techniques. the topics include: parallel N-Body Performance Comparison: Julia, Rust, and More;REFT: Resource-Efficient Federated Training Framework for Heterogeneous and Resource-Constrained Environments;An Efficient Data Provenance Collection Framework for HPC I/O Workloads;using Minicasts for Efficient Asynchronous Causal Unicast and Byzantine Tolerance;a Comparative Study of Two Matrix Multiplication algorithms Under Current Hardware architectures;Is Manual Code Optimization Still Required to Mitigate GPU thread Divergence? Applying a Flattening Technique to Observe Performance;towards Automatic, Predictable and High-Performance parallel Code Generation;Attack Graph Generation on HPC Clusters;analyzing the Influence of File Formats on I/O Patterns in Deep Learning;inference of Cell–Cell Interactions through Spatial Transcriptomics Data Using Graph Convolutional Neural Networks;natural Product-Like Compound Generation with Chemical Language Models;improved Early–Modern Japanese Printed Character Recognition Rate with Generated Characters;Improved Method for Similar Music Recommendation Using Spotify API;Reconfigurable Virtual Accelerator (ReVA) for Large-Scale Acceleration Circuits;Building Simulation Environment of Reconfigurable Virtual Accelerator (ReVA);vector Register Sharing Mechanism for High Performance Hardware Acceleration;Efficient Compute Resource Sharing of RISC-V Packed-SIMD Using Simultaneous Multi-threading;introducing Competitive Mechanism to Differential Evolution for Numerical Optimization;hyper-heuristic Differential Evolution with Novel Boundary Repair for Numerical Optimization;jump Like a Frog: Optimization of Renewable Energy Prediction in Smart Gird Based on Ultra Long Term Network;vision Transformer-Based Meta Loss Landscape Exploration with Actor-Critic Method;Fast Computation Method for Stopping Condition of Range Restricted
Massive multiple-input multiple-output (MIMO) detection plays a prominent role in the field of wireless communication. And withthis trend, message passing detection (MPD) attracts a great attention since its advantag...
Massive multiple-input multiple-output (MIMO) detection plays a prominent role in the field of wireless communication. And withthis trend, message passing detection (MPD) attracts a great attention since its advantages of high throughput and low complexity between distinct detection algorithms. Among various MPD algorithms, Gaussian approximate interference belief propagation (GAI-BP) detection can exhibit excellent convergence and performance in the different scenarios. However, its hardware design should be more efficient and flexible further. this paper proposes a novel semi-parallel hardware architecture for GAI-BP detection, aimed at improving its configurability using a layer scheduling scheme. through this scheme, the proposed detector achieves a gain of 0.65 dB when compared to the original algorithm, while simultaneously accommodating multi-antenna radios through the proposed hardware design.
A method for recovering the damaged uncompressed multichannel optical remote sensing data of the Earth is considered. Data represent the several frames of the same area, obtained from parallel turns, and were broken b...
详细信息
暂无评论