• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.12
• /
• pp.623-634
• /
• 2016

Most of existing frequent pattern mining methods address time efficiency and greatly rely on the primary memory. However, in the era of big data, the size of real-world databases to mined is exponentially increasing, and hence the primary memory is not sufficient enough to mine for frequent patterns from large real-world data sets. To solve this problem, there are some researches for frequent pattern mining method based on disk, but the processing time compared to the memory based methods took very time consuming. There are some researches to improve scalability of frequent pattern mining, but their processes are very time consuming compare to the memory based methods. In this paper, we present PPFP as a novel disk-based approach for mining frequent itemset from big data; and hence we reduced the main memory size bottleneck. PPFP algorithm is based on FP-growth method which is one of the most popular and efficient frequent pattern mining approaches. The mining with PPFP consists of two setps. (1) Constructing an IFP-tree: After construct FP-tree, we assign index number for each node in FP-tree with novel index numbering method, and then insert the indexed FP-tree (IFP-tree) into disk as IFP-table. (2) Mining frequent patterns with PPFP: Mine frequent patterns by expending patterns using stack based PUSH-POP method (PPFP method). Through this new approach, by using a very small amount of memory for recursive and time consuming operation in mining process, we improved the scalability and time efficiency of the frequent pattern mining. And the reported test results demonstrate them.

• Journal of KIISE
• /
• v.43 no.3
• /
• pp.304-313
• /
• 2016

We present a novel memory-efficient parallel ray casting algorithm for unstructured grid volume rendering on multi-core CPUs. Our method is based on the Bunyk ray casting algorithm. To solve the high memory overhead problem of the Bunyk algorithm, we allocate a fixed size local buffer for each thread and the local buffers contain information of recently visited faces. The stored information is used by other rays or replaced by other face's information. To improve the utilization of local buffers, we propose an image-plane based ray grouping algorithm that makes ray groups have high coherency. The ray groups are then distributed to computing threads and each thread processes the given groups independently. We also propose a novel hash function that uses the index of faces as keys for calculating the buffer index each face will use to store the information. To see the benefits of our method, we applied it to three unstructured grid datasets with different sizes and measured the performance. We found that our method requires just 6% of the memory space compared with the Bunyk algorithm for storing face information. Also it shows compatible performance with the Bunyk algorithm even though it uses less memory. In addition, our method achieves up to 22% higher performance for a large-scale unstructured grid dataset with less memory than Bunyk algorithm. These results show the robustness and efficiency of our method and it demonstrates that our method is suitable to volume rendering for a large-scale unstructured grid dataset.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.493-496
• /
• 2000

본 논문에서는 TMS320C6201 DSP (Digial Signal Processor)를 이용하여 실시간 영상 부호화기를 구현하였다. 기본적인 영상 압축 방법으로는 baseline-JPEG을 사용하였고 이에 움직임 검출 알고리즘을 부가하여 영상의 시간적인 중복성을 제거하였다. 특히 저속 메모리와 고속 메모리의 효율적인 분배 사용, 계산량이 많은 모듈의 최적화, 데이터의 병렬 연산과 DMA (Direct Memory Access)를 이용한 데이터 전송 등의 방법을 통하여 실시간 영상 부호화기의 고속 영상 처리에 중점을 두었다.

• Journal of Digital Contents Society
• /
• v.18 no.6
• /
• pp.1165-1173
• /
• 2017

Deep learning method, which is one of the most famous machine learning algorithms, has proven its applicability in various applications and is widely used in digital signal processing. However, it is difficult to apply deep learning technology to IoT devices with limited CPU performance and memory capacity, because a large number of training samples requires a lot of memory and computation time. In particular, if the Arduino with a very small memory capacity of 2K to 8K, is used, there are many limitations in implementing the algorithm. In this paper, we propose a method to optimize the ELM algorithm, which is proved to be accurate and efficient in various fields, on Arduino board. Experiments have shown that multi-class learning is possible up to 15-dimensional data on Arduino UNO with memory capacity of 2KB and possible up to 42-dimensional data on Arduino MEGA with memory capacity of 8KB. To evaluate the experiment, we proved the effectiveness of the proposed algorithm using the data sets generated using gaussian mixture modeling and the public UCI data sets.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.8
• /
• pp.377-384
• /
• 2007

With the advance of VLSI technology, the capacity and density of memories is rapidly growing. In this paper, We proposed a reallocation algorithm for faulty memory part to efficient reallocation with row and column redundant memory. Reallocation information obtained from faulty memory by only every test. Time overhead problem occurs geting reallocation information as every test. To its avoid, one test resulted from reallocation information can save to flash memory. In this paper, reallocation information increases efficiency using flash memory.

• Journal of KIISE:Software and Applications
• /
• v.31 no.4
• /
• pp.478-487
• /
• 2004

We had proposed the RPA(Recursive Partition Averaging) method in order to improve the storage requirement and classification rate of the Memory Based Reasoning. That algorithm worked not bad in many area, however, the major drawbacks of RPA are it's partitioning condition and the way of extracting major patterns. We propose an adaptive RPA algorithm which uses the FPD(feature-based population densimeter) to stop the ARPA partitioning process and produce, instead of RPA's averaged major pattern, optimizing resulting hyperrectangles. The proposed algorithm required only approximately 40% of memory space that is needed in k-NN classifier, and showed a superior classification performance to the RPA. Also, by reducing the number of stored patterns, it showed an excellent results in terms of classification when we compare it to the k-NN.

• The Journal of the Korea Contents Association
• /
• v.22 no.8
• /
• pp.81-93
• /
• 2022

Recently, research on processing a large-capacity graph using GPUs has been conducting. In order to process a large-capacity graph in a GPU with limited memory, the graph must be divided into subgraphs and then processed by scheduling subgraphs. In this paper, we propose an efficient graph algorithm processing scheme in GPU environments with limited memory and performance evaluation. The proposed scheme consists of a graph differential subgraph scheduling method and a graph segmentation method. The bulk graph segmentation method determines how a large-capacity graph can be segmented into subgraphs so that it can be processed efficiently by the GPU. The differential subgraph scheduling method schedule subgraphs processed by GPUs to reduce redundant transmission of the repeatedly used data between HOST-GPUs. It shows the superiority of the proposed scheme by performing various performance evaluations.

• Proceedings of the Korean Information Science Society Conference
• /
• 1999.10a
• /
• pp.60-62
• /
• 1999

SGML이나 XML과 같은 마크업 언어를 사용하여 생성된 구조적 문서에 대한 검색 시스템은 문서의 임의의 부분에 대한 검색을 지원한다. 문서의 구조에 바탕을 둔 다양한 유형의 사용자 질의를 처리하기 위해서는 색인에 필요한 메모리량이 커지게 된다. 색인에 필요한 메모리양을 줄이기 위해, 색인된 노드의 ID에서 찾고자 하는 노드의 ID를 계산할 수 있어야 한다. 그러나 이 경우 각 노드에 ID가 고정되기 때문에 문서의 갱신이 발생할 때 많은 부분이 수정되어야 하기 때문에 갱신에 필요한 오버헤드가 커지게 된다. 본 논문에서는 전체문서인스턴스트리 구조를 제안하고, 이를 기반으로 하여 노드의 ID를 구성함으로서, 색인과 검색의 효율성을 유지하면서 자료의 추가나 삭제등의 갱신이 발생할 때, 갱신의 파장을 최소화시킬 수 있는 색인구조와 질의처리 알고리즘을 제시한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.10d
• /
• pp.421-425
• /
• 2006

본 논문에서는 TinyOS 기반의 센서 네트워크에 대한 리프로그래밍을 에너지 효율적으로 수행하기 위한 매커니즘을 제안한다. 베이스 스테이션에서 센서노드에게 프로그램 전체를 보내는 대신 이전 버전과의 차이인 델타를 생성해서 전송할 때 Multi-round Rsync 알고리즘을 적용해 델타 파일의 크기를 최대한 줄이는 기법과 업데이트가 불필요한 플래시메모리 페이지에 대한 업데이트를 방지하기 위한 페이지 맵 기법을 통해 Rsync만을 사용하는 기존 방식보다 최대 30% 이상 에너지를 절감할 수 있음을 확인하였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.563-566
• /
• 2020

본 논문은 초고해상도 컴퓨터 홀로그램 생성을 위한 GPU 기반 2D Shift-FFT 의 효율적인 구현 방법을 제안한다. 본 연구가 제안하는 알고리즘은 기존에 여섯 단계로 이루어진 처리과정을 다섯 단계로 줄임으로서, 병렬처리에서 비효율적인 메모리 접근 과정을 줄인다. 또한, 핀드(pinned) 메모리 기반의 CPU-GPU 데이터 통신 통로인 핀드 버퍼(pinned buffer)를 사용하고 다중 스트림을 채용함으로써, GPU 활용의 주요 병목원인이 되는 데이터 통신의 부하를 줄이고 GPU 활용 효율을 높인다. 본 연구는 제안하는 알고리즘의 효용성을 증명하기 위해 서로 다른 두 시스템에 알고리즘을 구현하고, 다양한 크기의 행렬에 대한 2D-FFT 처리에 대한 성능을 측정하였다. 그 결과, CPU 기반의 FFTW 라이브러리 대비 최대 3 배, 동일한 GPU 를 사용하는 cuFFT 라이브러리 대비 최대 1.5 배 높은 성능을 달성하였다. 이러한 결과는, 본 연구가 제안하는 알고리즘의 효용성을 보여주는 결과다.