• Nuclear Engineering and Technology
• /
• v.55 no.7
• /
• pp.2516-2533
• /
• 2023

Several practical methods for accelerating the depletion calculation in a GPU-based Monte Carlo (MC) code PRAGMA are presented including the multilevel spectral collapse method and the vectorized Chebyshev rational approximation method (CRAM). Since the generation of microscopic reaction rates for each nuclide needed for the construction of the depletion matrix of the Bateman equation requires either enormous memory access or tremendous physical memory, both of which are quite burdensome on GPUs, a new method called multilevel spectral collapse is proposed which combines two types of spectra to generate microscopic reaction rates: an ultrafine spectrum for an entire fuel pin and coarser spectra for each depletion region. Errors in reaction rates introduced by this method are mitigated by a hybrid usage of direct online reaction rate tallies for several important fissile nuclides. The linear system to appear in the solution process adopting the CRAM is solved by the Gauss-Seidel method which can be easily vectorized on GPUs. With the accelerated depletion methods, only about 10% of MC calculation time is consumed for depletion, so an accurate full core cycle depletion calculation for a commercial power reactor (BEAVRS) can be done in 16 h with 24 consumer-grade GPUs.

• Smart Structures and Systems
• /
• v.32 no.3
• /
• pp.179-193
• /
• 2023

This study explores an alternative to the existing centralized process for data anomaly detection in modern Internet of Things (IoT)-based structural health monitoring (SHM) systems. An edge intelligence framework is proposed for the early detection and classification of various data anomalies facilitating quality enhancement of acquired data before transmitting to a central system. State-of-the-art deep neural network pruning techniques are investigated and compared aiming to significantly reduce the network size so that it can run efficiently on resource-constrained edge devices such as wireless smart sensors. Further, depthwise separable convolution (DSC) is invoked, the integration of which with advanced structural pruning methods exhibited superior compression capability. Last but not least, quantization-aware training (QAT) is adopted for faster processing and lower memory and power consumption. The proposed edge intelligence framework will eventually lead to reduced network overload and latency. This will enable intelligent self-adaptation strategies to be employed to timely deal with a faulty sensor, minimizing the wasteful use of power, memory, and other resources in wireless smart sensors, increasing efficiency, and reducing maintenance costs for modern smart SHM systems. This study presents a theoretical foundation for the proposed framework, the validation of which through actual field trials is a scope for future work.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.7
• /
• pp.2785-2799
• /
• 2020

We propose a novel speaker adaptation method using acoustic model clustering. The similarity of different speakers is defined by the cosine distance between their i-vectors (intermediate vectors), and various efficient clustering algorithms are applied to obtain a number of speaker subsets with different characteristics. The speaker-independent model is then retrained with the training data of the individual speaker subsets grouped by the clustering results, and an unknown speech is recognized by the retrained model of the closest cluster. The proposed method is applied to a large-scale speech recognition system implemented by a hybrid hidden Markov model and deep neural network framework. An experiment was conducted to evaluate the word error rates using Resource Management database. When the proposed speaker adaptation method using i-vector based clustering was applied, the performance, as compared to that of the conventional speaker-independent speech recognition model, was improved relatively by as much as 12.2% for the conventional fully neural network, and by as much as 10.5% for the bidirectional long short-term memory.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.12
• /
• pp.2675-2680
• /
• 2010

Computation steps for 3D graphic processing consist of two stages - fixed operation stage and programming required stage. Using this characteristic of 3D pipeline, a hybrid structure between graphics hardware designed by fixed structure and programmable hardware based on instructions, can handle graphic processing more efficiently. In this paper, fragment Shader is designed under this hybrid structure. It also supports OpenGL ES 2.0. Interior interface is optimized to reduce the delay of entire pipeline, which may be occurred by data I/O between the fixed hardware and the Shader. Interior register group of the Shader is designed by an interleaved structure to improve the register space and processing speed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.5
• /
• pp.1590-1609
• /
• 2021

Laughter is one of the most important nonverbal sound that human generates. It is a means for expressing his emotions. The acoustic and contextual features of this specific sound are different from those of speech and many difficulties arise during their modeling process. During this work, we propose an audio laughter generation system based on unsupervised generative models: the autoencoder (AE) and its variants. This procedure is the association of three main sub-process, (1) the analysis which consist of extracting the log magnitude spectrogram from the laughter database, (2) the generative models training, (3) the synthesis stage which incorporate the involvement of an intermediate mechanism: the vocoder. To improve the synthesis quality, we suggest two hybrid models (LSTM-VAE, GRU-VAE and CNN-VAE) that combine the representation learning capacity of variational autoencoder (VAE) with the temporal modelling ability of a long short-term memory RNN (LSTM) and the CNN ability to learn invariant features. To figure out the performance of our proposed audio laughter generation process, objective evaluation (RMSE) and a perceptual audio quality test (listening test) were conducted. According to these evaluation metrics, we can show that the GRU-VAE outperforms the other VAE models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.11
• /
• pp.5642-5670
• /
• 2017

Antivirus is an important issue to the security of virtual machine (VM). According to where the antivirus system resides, the existing approaches can be categorized into three classes: internal approach, external approach and hybrid approach. However, for the internal approach, it is susceptible to attacks and may cause antivirus storm and rollback vulnerability problems. On the other hand, for the external approach, the antivirus systems built upon virtual machine introspection (VMI) technology cannot find and prohibit viruses promptly. Although the hybrid approach performs virus scanning out of the virtual machine, it is still vulnerable to attacks since it completely depends on the agent and hooks to deliver events in the guest operating system. To solve the aforementioned problems, based on in-memory signature scanning, we propose an agentless runtime antivirus system VirtAV, which scans each piece of binary codes to execute in guest VMs on the VMM side to detect and prevent viruses. As an external approach, VirtAV does not rely on any hooks or agents in the guest OS, and exposes no attack surface to the outside world, so it guarantees the security of itself to the greatest extent. In addition, it solves the antivirus storm problem and the rollback vulnerability problem in virtualization environment. We implemented a prototype based on Qemu/KVM hypervisor and ClamAV antivirus engine. Experimental results demonstrate that VirtAV is able to detect both user-level and kernel-level virus programs inside Windows and Linux guest, no matter whether they are packed or not. From the performance aspect, the overhead of VirtAV on guest performance is acceptable. Especially, VirtAV has little impact on the performance of common desktop applications, such as video playing, web browsing and Microsoft Office series.

• Journal of Computing Science and Engineering
• /
• v.5 no.2
• /
• pp.121-130
• /
• 2011

Recent research indicates that transient errors will increasingly become a critical concern in microprocessor design. As embedded processors are widely used in reliability-critical or noisy environments, it is necessary to develop cost-effective fault-tolerant techniques to protect processors against transient errors. The register file is one of the critical components that can significantly affect microprocessor system reliability, since registers are typically accessed very frequently, and transient errors in registers can be easily propagated to functional units or the memory system, leading to silent data error (SDC) or system crash. This paper focuses on investigating the impact of register file soft errors on system reliability and developing cost-effective techniques to improve the register file immunity to soft errors. This paper proposes the register vulnerability factor (RVF) concept to characterize the probability that register transient errors can escape the register file and thus potentially affect system reliability. We propose an approach to compute the RVF based on register access patterns. In this paper, we also propose two compiler-directed techniques and a hybrid approach to improve register file reliability cost-effectively by lowering the RVF value. Our experiments indicate that on average, RVF can be reduced to 9.1% and 9.5% by the hyperblock-based instruction re-scheduling and the reliability-oriented register assignment respectively, which can potentially lower the reliability cost significantly, without sacrificing the register value integrity.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2009.11a
• /
• pp.160-160
• /
• 2009

In recent years, the merging of nanomaterials and nano-technology into electro-optic (EO) device technology such as liquid crystal displays (LCDs) has attracted much attention because of their unique electro- and magneto-optic properties and novel display applications. One example of hybrid LC-inorganic systems is semiconductor nanorods added to LC for their strong reorientation effect and tunable refractive index. Doping of nanoparticles in LC or polymers can lead to changes in performance characteristics such as electro-optical, dielectric, memory effect, phase behavior, etc. Due to the tunability of LCDs with mixed inorganic materials, low voltage operation of a LC system can also be achieved using the significant electro-optical effect achieved through suspension of ferroelectric nanoparticles in NLC.

• Proceedings of the Korean Information Science Society Conference
• /
• 2011.06a
• /
• pp.21-24
• /
• 2011

최근 NAND 플래시 메모리는 가벼운 무게, 적은 전력소모, 온도 및 충격에 강한 내구성 때문에 하드디스크를 대체할 저장 매체로 주목 받고 있다. 하지만 NAND 플래시 메모리는 비대칭적인 읽기 쓰기 소거 연산 처리 속도와 제자리 갱신이 불가능한 물리적인 특징으로 인해 디스크 기반의 대표적인 인덱스 구조 중의 하나인 해시 인덱스 구조를 NAND 플래시 메모리 상에 구현하였을 때, 레코드가 빈번하게 삽입, 삭제, 갱신되면 대량의 제자리 갱신이 발생하여 플래시 메모리에서 느린 쓰기 연산과 소거 연산이 수행되어 성능이 저하된다. 본 논문에서는 이러한 성능 저하를 피하기 위하여 버켓 오버플로우 발생 시 분할 연산을 수행하지 않고, 최대한 지연시킴으로써 쓰기 연산을 줄이는 인덱스 구조를 제안한다. 또한, 각 버켓에 대한 오버플로우 버켓의 갱신 및 삭제 비율에 따라 적응적으로 오버플로우 버켓을 할당하여 추가적인 읽기 쓰기 연산을 줄인다. 본 논문은 기존의 해시 인덱스 구조를 예제 및 수식을 통하여 제안하는 인덱스 구조의 우수성을 보인다.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.4
• /
• pp.50-56
• /
• 1997

In this paper, efficient quantization methods of line spectrum pairs (LSP) which has good performances and low complexity and memory are proosed for vocoder of mobile communication system. The adaptive quantization method utilizing the ordering property of LSP parameters is used in a scalar quantizer and a vector-scalar hybrid quantizer. The proposed scalar quantization algorithm needs 31 bits/frame to maintain the transparent quality of speech. The improved vector-scalar quantizer achieves an average spectral distortion of 1dB using 26 bits/frame. The proposed methods are evaluated in the channel errors and changed the predictor structure to maintain the robustness to channel errors.