检索结果-内蒙古大学图书馆

Hierarchical Recording Architecture for Three-Dimensional Magnetic Recording

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Jian, Yugen Luo, Ke Chen, Jincai Fong, Xuanyao Wuhan National Laboratory for Optoelectronics Huazhong University of Science and Technology Wuhan430074 China Department of Electrical and Computer Engineering National University of Singapore Singapore117583 Singapore School of Computer of Science and Technology Huazhong University of Science and Technology Wuhan430074 China Key Laboratory of Information Storage System Engineering Research Center of Data Storage Systems and Technology Ministry of Education of China Wuhan430074 China

Three-dimensional magnetic recording (3DMR) is a highly promising approach to achieving ultra-large data storage capacity in hard disk drives. One of the greatest challenges for 3DMR lies in performing sequential and correct writing of bits into the multi-layer recording medium. In this work, we have proposed a hierarchical recording architecture based on layered heat-assisted writing with a multi-head array. The feasibility of the architecture is validated in a dual-layer 3DMR system with FePt-based thin films via micromagnetic simulation. Our results reveal the magnetization reversal mechanism of the grains, ultimately attaining appreciable switching probability and medium signal-to-noise ratio (SNR) for each layer. In particular, an optimal head-to-head distance is identified as the one that maximizes the medium SNR. Optimizing the system’s noise resistance will improve the overall SNR and allow for a smaller optimal head-to-head distance, which can pave the way for scaling 3DMR to more recording layers. Copyright © 2025, The Authors. All rights reserved.

关键词： Hard disk storage

Evolutionary Approach to Security Games with Signaling

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Żychowski, Adam Mańdziuk, Jacek Bondi, Elizabeth Venugopal, Aravind Tambe, Milind Ravindran, Balaraman Center for Research on Computation and Society Harvard University United States Faculty of Mathematics and Information Science Warsaw University of Technology Poland Robert Bosch Centre for Data Science and AI IIT Madras India Department of Computer Science and Engineering IIT Madras India

Green Security Games have become a popular way to model scenarios involving the protection of natural resources, such as wildlife. Sensors (e.g. drones equipped with cameras) have also begun to play a role in these scenarios by providing real-time information. Incorporating both human and sensor defender resources strategically is the subject of recent work on Security Games with Signaling (SGS). However, current methods to solve SGS do not scale well in terms of time or memory. We therefore propose a novel approach to SGS, which, for the first time in this domain, employs an Evolutionary Computation paradigm: EASGS. EASGS effectively searches the huge SGS solution space via suitable solution encoding in a chromosome and a specially-designed set of operators. The operators include three types of mutations, each focusing on a particular aspect of the SGS solution, optimized crossover and a local coverage improvement scheme (a memetic aspect of EASGS). We also introduce a new set of benchmark games, based on dense or locally-dense graphs that reflect real-world SGS settings. In the majority of 342 test game instances, EASGS outperforms state-of-the-art methods, including a reinforcement learning method, in terms of time scalability, nearly constant memory utilization, and quality of the returned defender’s strategies (expected payoffs). © 2022, CC BY.

关键词： Reinforcement learning

Exploring Visibility-Based Shading Methods for Graphics and Visualization

学校读者我要写书评

暂无评论

Exploring Visibility-Based Shading Methods for Graphics and ...

International Conference on Computing and Networking Technology (ICCNT)

作者： D. Anandhasilambarasan Kanika Seth Ar Aravind Prakhar Goyal I K Kantharaj Yashwant S. Ingle Department of Computer Science Engineering Karpagam Academy of Higher Education Coimbatore India Centre of Research Impact and Outcome Chitkara University Rajpura Punjab India Department of Electronics and Communication Engineering Prince Shri Venkateshwara Padmavathy Engineering College Chennai India Quantum University Research Center Quantum University School of Engineering and Technology Mechaical Engineering JAIN (Deemed to be University) Bangalore Karnataka India Department of Artificial Intelligence & Data Science Vishwakarma Institute of Technology Pune India

ISBN: (数字)9798350370249

ISBN: (纸本)9798350370270

Visibility-based shading techniques for laptop snapshots and Visualization can provide realistic and perceptual shading consequences with the aid of utilizing data from the scene’s visibility shape. Those methods comprise visibility statistics to compute the approximate illumination of an object or scene, thinking about occlusion consequences from other objects. Visibility-based totally shading algorithms permit more practical actual-time renderings that extra accurately fit the way people perceive the illumination of scenes. Moreover, these techniques offer performance blessings over global illumination-based total methods, such as improved memory burden and faster computation instances, making them appropriate for plenty of applications. Popular visibility-based strategies, including ray casting and direct illumination, are based totally on the idea that each surface is ideally diffuse and homogeneous, which means their absorption and reflection coefficients are constant. To overcome this issue, a range of new techniques integrate visibility and physics-based total fashions, accounting for fabric homes (such as specular highlights) in the illumination computation. This combination of visibility and physics-based fashions can be, in addition, improved via incorporating mild shipping simulation strategies, inclusive of bidirectional reflectance distribution features and course tracing, to capture better and predict the rendering of actual international materials and surfaces.

关键词： Visualization Portable computers Shape Lighting Ray tracing Rendering (computer graphics) Real-time systems Fabrics Reflection coefficient Photonics

Tsunami Evacuation Model Assessment in the Panimbang Subdistrict, Banten Province, Indonesia: GIS- and Agent-Based Modeling Approaches

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Purbani, Dini Marzuki, Marza Ihsan Ontowirjo, Budianto Zein, Farhan Makarim Wisha, Ulung Jantama Research Center for Conservation of Marine and Inland Resources National Research and Innovation Agency Cibinong Indonesia Research Center for Science Data and Information National Research and Innovation Agency Bandung Indonesia Civil Engineering Department Faculty of Engineering and Computer Science Bakrie University Jakarta Indonesia Department of Geography Faculty of Mathematics and Natural Sciences University of Indonesia Depok Indonesia Research Center for Oceanography National Research and Innovation Agency Jakarta Indonesia Department of Physics and Earth Sciences Graduate School of Engineering and Science University of the Ryukyus Nishihara Japan

This study examines the agent behavior during a complex tsunami process to reach designated shelters based on the agent-based modeling approach. A GIS-based tsunami inundation modelling is employed considering the land slope and surface roughness coefficient to estimate the tsunami height loss. This model becomes the basis for determining the service area within the study site. Five shelter candidates are assessed using agent-based modeling simulated in the NetLogo, implementing the 40 minutes ETA and 9-m tsunami runup propagation for four population age classifications. Overall, the Panimbang coastal area is categorized as high vulnerability to tsunamis (14.68% out of the total area), where the runup propagation is extended due to the tsunami invading rivers, thereby increasing the risk toward local society. Of particular concern, two proposed shelters (SA1 and SA5) exceed the capacity by 12.43% and 59.44% out of their maximum capacity, respectively. Of 9640 agents (people) simulated in the NetLogo, 77.7% are evacuated, and 22.3% fail to reach the shelters, predominated by the adult category. A sufficiently high casualty number is due to a "bottleneck" in the overlapping service areas. However, reconsidering the overcapacity TES with some additional shelters in the surrounding area is possible. © 2023, The Authors. All rights reserved.

关键词： Tsunamis

No Fear of Representation Bias Graph Contrastive Learning with Calibration and Fusion

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Li, Jindong Liu, Yixin Xing, Qianli Wang, Qi Pan, Shirui The School of Artificial Intelligence Jilin University Changchun130012 China The Department of Data Science and AI Monash University MelbourneVIC3800 Australia The College of Computer Science and Technology Jilin University Changchun130012 China The School of Information and Communication Technology Griffith University Gold CoastQLD4222 Australia Engineering Research Center of Knowledge-Driven Human-Machine Intelligence Ministry of Education China

Graph contrastive learning (GCL) is receiving growing attention due to its promising self-supervised learning capability on graph-structured data. While GCL approaches have demonstrated significant potential, their inherited stochastic perturbation-based augmentation causes the representation bias on representation hyperspace. Such a serious issue, unfortunately, causes the representation drift between the distributions of original and augmented embeddings, misleading the contrastive training for the graph neural network encoders. Moreover, the information loss caused by bias can further result in incomplete representation problem, hindering the self-supervised framework from capturing comprehensive structural and attributive patterns. To address these challenges, this paper introduces the CAlibratioN aNd FusiON model (CANNON for short). CANNON introduces a cross-view calibrating contrast mechanism to enhance the agreement between the original and augmented representations, which effectively alleviates the representation drift. Concurrently, CANNON overcomes the incomplete representation problem with a fused-view calibrating contrast mechanism for contrastive learning application on the fused embeddings, where a transformer-based cross-view fusion block is used to integrate two augmented views into a comprehensive one. The extensive experimental results, executed on 9 real-world datasets, demonstrate that CANNON significantly outperforms existing state-of-the-art methods. The code is available at https://***/jindongli-ai/CANNON. © 2024, The Authors. All rights reserved.

关键词： Calibration

IDENTIFYING SOURCE SPEAKERS FOR VOICE CONVERSION BASED SPOOFING ATTACKS ON SPEAKER VERIFICATION SYSTEMS

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Cai, Danwei Cai, Zexin Li, Ming Department of Electrical and Computer Engineering Duke University Durham United States Data Science Research Center Duke Kunshan University Kunshan China

An automatic speaker verification system aims to verify the speaker identity of a speech signal. However, a voice conversion system could manipulate a person's speech signal to make it sound like another speaker's voice and deceive the speaker verification system. Most countermeasures for voice conversion-based spoofing attacks are designed to discriminate bona fide speech from spoofed speech for speaker verification systems. In this paper, we investigate the problem of source speaker identification - inferring the identity of the source speaker given the voice converted speech. To perform source speaker identification, we simply add voice-converted speech data with the label of source speaker identity to the genuine speech dataset during speaker embedding network training. Experimental results show the feasibility of source speaker identification when training and testing with converted speeches from the same voice conversion model(s). In addition, our results demonstrate that having more converted utterances from various voice conversion model for training helps improve the source speaker identification performance on converted utterances from unseen voice conversion models. © 2022, CC BY.

关键词： Speech recognition

Saliencycut: Augmenting Plausible Anomalies for Anomaly Detection

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Ye, Jianan Hu, Yijie Yang, Xi Wang, Qiu-Feng Huang, Chao Huang, Kaizhu School of Advanced Technology Xi’an Jiaotong-Liverpool University China Department of Electrical Engineering and Electronics University of Liverpool United Kingdom Department of Computer Science University of Liverpool United Kingdom Data Science Research Center Duke Kunshan University China

Anomaly detection under the open-set scenario is a challenging task that requires learning discriminative features to detect anomalies that were even unseen during training. As a cheap yet effective approach, data augmentation has been widely used to create pseudo anomalies for better training of such models. Recent wisdom of augmentation methods focuses on generating random pseudo instances that may lead to a mixture of augmented instances with seen anomalies, or out of the typical range of anomalies. To address this issue, we propose a novel saliency-guided data augmentation method, SaliencyCut, to produce pseudo but more common anomalies that tend to stay in the plausible range of anomalies. Furthermore, we deploy a two-head learning strategy consisting of normal and anomaly learning heads to learn the anomaly score of each sample. Theoretical analyses show that this mechanism offers a more tractable and tighter lower bound of the data log-likelihood. We then design a novel patch-wise residual module in the anomaly learning head to extract and assess anomaly features from each sample, facilitating the learning of discriminative representations of anomaly instances. Extensive experiments conducted on six real-world anomaly detection datasets demonstrate the superiority of our method to competing methods under various settings. Code is available at: https://***/yjnanan/SaliencyCut. © 2023, The Authors. All rights reserved.

关键词： Anomaly detection

Prediction of Stock Price data of PT. Ramayana Lestari Sentosa Tbk. using Long Short Term Memory Model

学校读者我要写书评

暂无评论

Prediction of Stock Price Data of PT. Ramayana Lestari Sento...

2021 International Conference on Artificial Intelligence and Big data Analytics, ICaiBDA 2021

作者： Setiawan, Luki Muntasir, Nurul Fathanah Fahreza, Syafa Sholahuddin, Asep Universitas Padjadjaran Research Center for Artificial Intelligence and Big Data Indonesia Department of Mathematics Universitas Padjadjaran Sumedang Indonesia Department of Electrical Engineering Hasanuddin University Makassar Indonesia Department of Computer Science Telkom University Bandung Indonesia Department of Computer Science Universitas Padjadjaran Sumedang Indonesia

ISBN: (纸本)9781665408905

The investment that Indonesian people are interested in during the pandemic is investing in stocks. However, stock investment will not only generate profits but also risk losses so that to minimize the risk of loss, forecasting is required. One of the prediction methods that can be used is Long Short Term Memory (LSTM) for stock data modeling, this is because LSTM is suitable for predicting and processing important events with relatively long intervals and delays in the time series data. The data used in this research is time series data of PT. Ramayana Lestari Sentosa Tbk. in the period 2 March 2020 to 15 September 2021 with LSTM model using the same number of epochs but different number of nodes, activation functions and optimizers which show varying results and accuracy levels. The selection of the number of epochs 5 is the optimal number of epochs selected. In the experiment, the researchers use 2 types of nodes (4 and 16), 4 types of activation functions (Linear, Sigmoid, Relu, and Tanh), and 3 types of optimizers (Adam, SGD, and Adagrad). The results of the experiment show that the best LSTM model based on the smallest RMSE test value is 15.37 on the use of LSTM model with 4 nodes, Linear activation function, and Adam optimizer. Meanwhile, the stock price prediction for the next day (16 September 2021) obtained based on LSTM model is IDR 639.05. Based on this, it can be concluded that in predicting RALS stock price data using LSTM model with the number of nodes 4 is better than the number of nodes 16. Linear activation function is also better than the activation function of Sigmoid, Relu and Tanh. Then, Adam optimizer is better than SGD and Adagrad optimizer. © 2021 IEEE.

关键词： Forecasting

The DKU-DukeECE Diarization System for the VoxCeleb Speaker Recognition Challenge 2022

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wang, Weiqing Qin, Xiaoyi Cheng, Ming Zhang, Yucong Wang, Kangyue Li, Ming Data Science Research Center Duke Kunshan University Kunshan China Department of Electrical and Computer Engineering Duke University Durham United States

This paper discribes the DKU-DukeECE submission to the 4th track of the VoxCeleb Speaker Recognition Challenge 2022 (VoxSRC-22). Our system contains a fused voice activity detection model, a clustering-based diarization model, and a target-speaker voice activity detection-based overlap detection model. Overall, the submitted system is similar to our previous year's system in VoxSRC-21. The difference is that we use a much better speaker embedding and a fused voice activity detection, which significantly improves the performance. Finally, we fuse 4 different systems using DOVER-lap and achieve 4.75% of the diarization error rate, which ranks the 1st place in track 4. Copyright © 2022, The Authors. All rights reserved.

关键词： Speech recognition