• Journal of KIISE:Computer Systems and Theory
• /
• v.28 no.5
• /
• pp.228-235
• /
• 2001

본 논문에서는 분산메모리를 가진 병렬 컴퓨터에서 밀집 행렬 연산을 위한 PoLAPACK 패키지를 소개한다. PoLAPACK은 새로운 연산 기법을 적용한 LU, QR, Cholesky 인수분해 알고리즘들을 포함하고 있다. 블록순환분산법으로 분산되어 있는 행렬에 알고리즘적인 블록 기법(algorithimic blocking)을 적용하여, 실제 행렬의 분산에 사용된 블록의 크기와 다른, 최대의 성능을 보일 수 있는 최적의 블록 크기로 연산을 수행할 수 있다. 이러한 연산 방식은 분산되어 있는 원래의 행렬 A의 순서를 따르지 않으며, 따라서 최적의 블록 크기로 연산을 수행한 후에 얻어진 해 x를 원래 행렬 분산법을 따라서 재배치하여야 한다. 본 연구는 Cray T3E 컴퓨터에서 구현하였으며 ScaLAPACK의 인수분해 루틴들과 그 성능을 비교.분석하였다.

• The KIPS Transactions:PartB
• /
• v.13B no.6 s.109
• /
• pp.625-632
• /
• 2006

Collaborative filtering is a technology that aims at teaming predictive models of user preferences. Collaborative filtering systems have succeeded in Ecommerce market but they have shortcomings of high dimensionality and sparsity. In this paper we propose the nearest neighbor collaborative filtering method using non-negative matrix factorization(NNMF). We replace the missing values in the user-item matrix by using the user variance coefficient method as preprocessing for matrix decomposition and apply non-negative factorization to the matrix. The positive decomposition method using the non-negative decomposition represents users as semantic vectors and classifies the users into groups based on semantic relations. We compute the similarity between users by using vector similarity and selects the nearest neighbors based on the similarity. We predict the missing values of items that didn't rate by a new user based on the values that the nearest neighbors rated items.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2017.11a
• /
• pp.202-203
• /
• 2017

비음수 행렬 인수분해(Non-negative Matrix Factorization, NMF)의 신호분리 성능을 개선하기 위해 희소조건을 인가한 방법이 희소 비음수 행렬 인수분해 알고리즘(Sparse NMF, SNMF)이다. 기존의 SNMF 알고리즘은 개별 음원의 희소성을 고려하지 않고 임의로 결정한 희소 조건을 사용한다. 본 논문에서는 음원의 특성에 따른 희소성을 추정하고 이를 SNMF 학습알고리즘에 적용하는 새로운 신호분리 기법을 제안한다. 혼합 신호에서의 잡음제거 실험을 통해, 제안한 방법이 기존의 NMF와 SNMF에 비해 성능이 더 우수함을 보였다.

• Journal of the Korea Computer Industry Society
• /
• v.7 no.4
• /
• pp.293-298
• /
• 2006

A factorization is an extremely important part of multi-level logic synthesis. The number of literals in a factored form is a good estimate of the complexity of a logic function, and can be translated directly into the number of transistors required for implementation. Factored forms are described as either algebraic or Boolean, according to the trade-off between run-time and optimization. A Boolean factored form contains fewer number of literals than an algebraic factored form. In this paper, we present a new method for a Boolean factorization. The key idea is to identify two-cube Boolean subexpression pairs from given expression. Experimental results on various benchmark circuits show the improvements in literal counts over the algebraic factorization based on Bryton's co-kernel cube matrix.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.8C
• /
• pp.661-666
• /
• 2010

In this paper, we apply a likelihood ratio test (LRT) to a non-negative matrix factorization (NMF) based voice activity detection (VAD) to find optimal threshold. In our approach, the NMF based VAD is expressed as Euclidean distance between noise basis vector and input basis vector which are extracted through NMF. The optimal threshold each of noise environments depend on NMF results distribution in noise region which is estimated statistical model-based VAD. According to the experimental results, the proposed approach is found to be effective for statistical model-based VAD using LRT.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.10
• /
• pp.1923-1928
• /
• 2017

This paper is a study on error correction for three-dimensional shape reconstruction based on factorization method. The existing error correction method based on factorization has a limitation of correction because it is optimized globally. Thus in this paper, we propose our new method which can find and correct the only major error influence factor toward three-dimensional reconstructed shape instead of global approach. We define the error-influenced factor in two-dimensional re-projection deviation space and directly control the error components. In addition, it is possible to improve the error correcting performance by recursively applying the above process. This approach has an advantage under noise because it controls the major error components without depending on any geometric information. The performance evaluation of the proposed algorithm is verified by simulation with synthetic and real image sequence to demonstrate noise robustness.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.4 s.316
• /
• pp.1-7
• /
• 2007

In this paper, we analyze geometrical relations of 3D shape reconstruction from 2D images taken under anne projection. The purpose of this research is to contribute to more accurate 3-D reconstruction under noise distribution by analyzing geometrically the 2D to 3D relationship. In situation for no missing feature points (FPs) or no noise in 2D image plane, the accurate solution of 3D shape reconstruction is blown to be provided by Singular Yalue Decomposition (SVD) factorization. However, if several FPs not been observed because of object occlusion and image low resolution, and so on, there is no simple solution. Moreover, the 3D shape reconstructed from noise-distributed FPs is peturbed because of the influence of the noise. This paper focuses on analysis of geometrical properties which can interpret the missing FPs even though the noise is distributed on other FPs.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.4
• /
• pp.959-966
• /
• 1998

We present fast QR factorization algorithms $m{\times}n\;(m{\geq}n)$ Toeplitz matrix. These QR factorization algortihms are determined from the shift-invariance properties of underlying matrices. The major transformation tool is a stabilized/hyperbolic Householder transformation. The algortihms require O(mn) operations, and can be easily implemented on distributed-memory multiprocessors.

• Journal of the Korea Computer Industry Society
• /
• v.8 no.4
• /
• pp.229-236
• /
• 2007

A factored form is a sum of products of sums of products, ..., of arbitrary depth. Factoring is the process of deriving a parenthesized form with the smallest number of literals from a two-level form of a logic expression. The factored form is not unique and described as either algebraic or Boolean. A Boolean factored form contains fewer number of literals than an algebraic factored form. In this paper, we present a new method for a Boolean factorization. The key idea is to identify two-cube Boolean subexpressions from given two-level logic expression and to extract divisor/quotient pairs. Then, we derive extended divisor/quotient pairs, where their quotients are not cube-free, from the generated divisor/quotients pairs. We generate quotient/quotient pairs from divisor/quotient pairs and extended divisor/quotient pairs. Using the pairs, we make a matrix to generate Boolean factored form based on a technique of rectangle covering.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2006.05a
• /
• pp.469-472
• /
• 2006

본 논문은 비음수 행렬 인수분해(NMF, non-negative matrix factorization)를 이용하여 문장을 추출하여 문서를 요약하는 새로운 방법을 제안하였다. 제안된 방법은 문장추출에 사용되는 의미 특징(semantic feature)이 비 음수 값을 갖기 때문에 잠재의미분석에 비해 문서의 내용을 정확하게 요약한다. 또한, 적은 계산비용을 통하여 쉽게 요약 문장을 추출할 수 있는 장점을 갖는다.