1-6hit |
Wenbing JIN Xuanya LI Yanyong YU Yongzhi WANG
To improve Last-Level Cache (LLC) management, numerous approaches have been proposed requiring additional hardware budget and increased overhead. A number of these approaches even change the organization of the existing cache design. In this letter, we propose Adaptive Insertion and Promotion (AIP) policies based on Least Recently Used (LRU) replacement. AIP dynamically inserts a missed line in the middle of the cache list and promotes a reused line several steps left, realizing the combination of LRU and LFU policies deliberately under a single unified scheme. As a result, it benefits workloads with high locality as well as with many frequently reused lines. Most importantly, AIP requires no additional hardware other than a typical LRU list, thus it can be easily integrated into the existing hardware with minimal changes. Other issues around LLC such as scans, thrashing and dead lines are all explored in our study. Experimental results on the gem5 simulator with SPEC CUP2006 benchmarks indicate that AIP outperforms LRU replacement policy by an average of 5.8% on the misses per thousand instructions metric.
Yesheng GAO Hui SHENG Kaizhi WANG Xingzhao LIU
A signal-model-based SAR image formation algorithm is proposed in this paper. A model is used to describe the received signal, and each scatterer can be characterized by a set of its parameters. Two parameter estimation methods via atomic decomposition are presented: (1) applying 1-D matching pursuit to azimuthal projection data; (2) applying 2-D matching pursuit to raw data. The estimated parameters are mapped to form a SAR image, and the mapping procedure can be implemented under application guidelines. This algorithm requires no prior information about the relative motion between the platform and the target. The Cramer-Rao bounds of parameter estimation are derived, and the root mean square errors of the estimates are close to the bounds. Experimental results are given to validate the algorithm and indicate its potential applications.
Hongwei HAN Ke GUO Maozhi WANG Tingbin ZHANG Shuang ZHANG
The sparse unmixing of hyperspectral data has attracted much attention in recent years because it does not need to estimate the number of endmembers nor consider the lack of pure pixels in a given hyperspectral scene. However, the high mutual coherence of spectral libraries strongly affects the practicality of sparse unmixing. The collaborative sparse unmixing via variable splitting and augmented Lagrangian (CLSUnSAL) algorithm is a classic sparse unmixing algorithm that performs better than other sparse unmixing methods. In this paper, we propose a CLSUnSAL-based hyperspectral unmixing method based on dictionary pruning and reweighted sparse regression. First, the algorithm identifies a subset of the original library elements using a dictionary pruning strategy. Second, we present a weighted sparse regression algorithm based on CLSUnSAL to further enhance the sparsity of endmember spectra in a given library. Third, we apply the weighted sparse regression algorithm on the pruned spectral library. The effectiveness of the proposed algorithm is demonstrated on both simulated and real hyperspectral datasets. For simulated data cubes (DC1, DC2 and DC3), the number of the pruned spectral library elements is reduced by at least 94% and the runtime of the proposed algorithm is less than 10% of that of CLSUnSAL. For simulated DC4 and DC5, the runtime of the proposed algorithm is less than 15% of that of CLSUnSAL. For the real hyperspectral datasets, the pruned spectral library successfully reduces the original dictionary size by 76% and the runtime of the proposed algorithm is 11.21% of that of CLSUnSAL. These experimental results show that our proposed algorithm not only substantially improves the accuracy of unmixing solutions but is also much faster than some other state-of-the-art sparse unmixing algorithms.
Kanghui ZHAO Tao LU Yanduo ZHANG Yu WANG Yuanzhi WANG
In recent years, compared with the traditional face super-resolution (SR) algorithm, the face SR based on deep neural network has shown strong performance. Among these methods, attention mechanism has been widely used in face SR because of its strong feature expression ability. However, the existing attention-based face SR methods can not fully mine the missing pixel information of low-resolution (LR) face images (structural prior). And they only consider a single attention mechanism to take advantage of the structure of the face. The use of multi-attention could help to enhance feature representation. In order to solve this problem, we first propose a new pixel attention mechanism, which can recover the structural details of lost pixels. Then, we design an attention fusion module to better integrate the different characteristics of triple attention. Experimental results on FFHQ data sets show that this method is superior to the existing face SR methods based on deep neural network.
Juan CHEN Shen SU Xianzhi WANG
Location sharing services have recently gained momentum over mobile online social networks (mOSNs), seeing the increasing popularity of GPS-capable mobile devices such as smart phones. Despite the convenience brought by location sharing, there comes severe privacy risks. Though many efforts have been made to protect user privacy during location sharing, many of them rely on the extensive deployment of trusted Cellular Towers (CTs) and some incur excessive time overhead. More importantly, little research so far can support complete privacy including location privacy, identity privacy and social relation privacy. We propose SAM, a new System Architecture for mOSNs, and P3S, a Privacy-Preserving Protocol based on SAM, to address the above issues for privacy-preserving location sharing over mOSNs. SAM and P3S differ from previous work in providing complete privacy for location sharing services over mOSNs. Theoretical analysis and extensive experimental results demonstrate the feasibility and efficiency of the proposed system and protocol.
Yingwei FU Kele XU Haibo MI Qiuqiang KONG Dezhi WANG Huaimin WANG Tie HONG
Sound event detection is intended to identify the sound events in audio recordings, which has widespread applications in real life. Recently, convolutional recurrent neural network (CRNN) models have achieved state-of-the-art performance in this task due to their capabilities in learning the representative features. However, the CRNN models are of high complexities with millions of parameters to be trained, which limits their usage for the mobile and embedded devices with limited computation resource. Model distillation is effective to distill the knowledge of a complex model to a smaller one, which can be deployed on the devices with limited computational power. In this letter, we propose a novel multi model-based distillation approach for sound event detection by making use of the knowledge from models of multiple teachers which are complementary in detecting sound events. Extensive experimental results demonstrated that our approach achieves a compression ratio about 50 times. In addition, better performance is obtained for the sound event detection task.