• ETRI Journal
• /
• 제32권5호
• /
• pp.795-800
• /
• 2010

Embedding a large vocabulary speech recognition system in mobile devices requires a reduced acoustic model obtained by eliminating redundant model parameters. In conventional optimization methods based on the minimum description length (MDL) criterion, a binary Gaussian tree is built at each state of a hidden Markov model by iteratively finding and merging similar mixture components. An optimal subset of the tree nodes is then selected to generate a downsized acoustic model. To obtain a better binary Gaussian tree by improving the process of finding the most similar Gaussian components, this paper proposes a new distance measure that exploits the difference in likelihood values for cases before and after two components are combined. The mixture weight of Gaussian components is also introduced in the component merging step. Experimental results show that the proposed method outperforms MDL-based optimization using either a Kullback-Leibler (KL) divergence or weighted KL divergence measure. The proposed method could also reduce the acoustic model size by 50% with less than a 1.5% increase in error rate compared to a baseline system.

• IEIE Transactions on Smart Processing and Computing
• /
• 제5권5호
• /
• pp.331-336
• /
• 2016

Multi-bit tries have been proposed to improve the search performance of a binary trie by providing flexibility in stride values, which identify the number of bits examined at a time. However, constructing a variable-stride multi-bit trie is challenging since it is not easy to determine a proper stride value that satisfies the required performance at each node. The aim of this paper is to identify several desired characteristics of a trie for IP address lookup problems, and to propose a multi-stride decision trie that has these characteristics. Simulation results using actual routing sets with about 30,000 to 220,000 prefixes show that the proposed multi-stride decision trie has the desired characteristics and achieves IP address lookup using 33% to 47% of the 2-bit trie in the average number of node accesses, while requiring a smaller amount of memory.

• Nuclear Engineering and Technology
• /
• 제40권7호
• /
• pp.571-580
• /
• 2008

A Binary Decision Diagram (BDD) is a graph-based data structure that calculates an exact top event probability (TEP). It has been a very difficult task to develop an efficient BDD algorithm that can solve a large problem since it is highly memory consuming. In order to solve a large reliability problem within limited computational resources, many attempts have been made, such as static and dynamic variable ordering schemes, to minimize BDD size. Additional effort was the development of a ZBDD (Zero-suppressed BDD) algorithm to calculate an approximate TEP. The present method is the first successful application of a BDD truncation. The new method is an efficient method to maintain a small BDD size by a BDD truncation during a BDD calculation. The benchmark tests demonstrate the efficiency of the developed method. The TEP rapidly converges to an exact value according to a lowered truncation limit.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2002년도 추계학술대회 및 정기총회
• /
• pp.383-386
• /
• 2002

In this paper, we investigated the pattern recognition performance of the numeric patterns (from 0 to 9) using DNA coding method. The pattern recognition performance of the DNA coding method is compared to the that of the GA(Genetic Algorithm). GA searches effectively an optimal solution via the artificial evolution of individual group of binary string using binary coding, while DNA coding method uses four-type bases denoted by A(Adenine), C(Cytosine), G(Guanine) and T(Thymine), The pattern recognition performance of GA and DNA coding method is evaluated by using the same genetic operators(crossover and mutation) and the crossover probability and mutation probability are set the same value to the both methods. The DNA coding method has better characteristics over genetic algorithms (GA). The reasons for this outstanding performance is multiple possible solution presentation in one string and variable solution string length.

• 품질경영학회지
• /
• 제26권1호
• /
• pp.143-160
• /
• 1998

This article is concerned with the selection of subsets of predictor variables to be included in bulding the binary response logistic regression model. It is based on a Bayesian aproach, intended to propose and develop a procedure that uses probabilistic considerations for selecting promising subsets. This procedure reformulates the logistic regression setup in a hierarchical normal mixture model by introducing a set of hyperparameters that will be used to identify subset choices. It is done by use of the fact that cdf of logistic distribution is a, pp.oximately equivalent to that of $t_{(8)}$/.634 distribution. The a, pp.opriate posterior probability of each subset of predictor variables is obtained by the Gibbs sampler, which samples indirectly from the multinomial posterior distribution on the set of possible subset choices. Thus, in this procedure, the most promising subset of predictors can be identified as that with highest posterior probability. To highlight the merit of this procedure a couple of illustrative numerical examples are given.

• Communications for Statistical Applications and Methods
• /
• 제29권3호
• /
• pp.319-331
• /
• 2022

In this paper, we develop a new statistical model to forecast the PM_2.5 level in Seoul, South Korea. The proposed model is based on the extreme quantile regression model with lasso penalty. Various meteorological variables and air pollution variables are considered as predictors in the regression model, and the lasso quantile regression performs variable selection and solves the multicollinearity problem. The final prediction model is obtained by combining various extreme lasso quantile regression estimators and we construct a binary classifier based on the model. Prediction performance is evaluated through the statistical measures of the performance of a binary classification test. We observe that the proposed method works better compared to the other classification methods, and predicts 'very bad' cases of the PM_2.5 level well.

• Journal of information and communication convergence engineering
• /
• 제21권1호
• /
• pp.68-74
• /
• 2023

This study examines the response of users' feelings from a gender perspective toward interactive video mobile learning (IVML). An IVML prototype was developed for the Android platform allowing users to install and make use of the app for m-learning purposes. This study aims to measure the level of feelings toward the IVML prototype and examine the differences in gender perspectives, identify the most responsive feelings between male, and female users as prominent feelings and measure the correlation between user-friendly feeling traits as an independent variable in accordance with gender attributes. The feelings response could then be extracted from the user experience, user interface, and human-computer interaction based on gender perspectives using the Kansei engineering approach as the measurement method. The statistical results demonstrated the different emotional reactions from a male and female perspective toward the IVML prototype may or may not have a correlation with the user-friendly trait, perhaps having a similar emotional response from one to another.

• Journal of the Korean Statistical Society
• /
• 제6권2호
• /
• pp.117-154
• /
• 1977

본논문에서는 첫째로 단일방정식 모형에서 종속변수가 양자택일(binary choice)의 이산확률변수일 때 이러한 이진적 종속변수(binary dependent variable)의 변동을 설명하는데 사용되는 몇 가지 모형을 소개하고 각각의 표기 및 추정방법, 추정량의 성질, 예측 및 검정 문제 등에 관하여 비교 서술하고자 한다. 둘째, 종속변수가 이산과 연속의 혼합형태일 때 앞에 소개된 모형이 어떻게 적용될 수 있는가를 살펴보며, 셋째, 선택대상 및 종속변수의 수가 증가하여 일반화된 선다형모형(multiple choice model)의 경우, 표기 및 추정방법을 단일방정식 기법을 이용하여 추가로 총람하고자 한다. 넷째, 본논문에서는 또한 내구소비재 구입에 관한 조사자료를 이용하여 실제 많이 사용되는 몇 개의 모형을 선택하여 적용하고 각각의 예측력을 분석함으로써 각 모형을 비교 검토하는데 목적이 있다.

• 한국정보처리학회논문지
• /
• 제6권3호
• /
• pp.781-791
• /
• 1999

DD(Decision Diagrams) is an efficient operational data structure for an optimal expression of boolean functions. In a graph-based synthesis using DD, the goal of optimization decreases representation space for boolean functions. This paper represents boolean functions using OPKFDD(Ordered Pseudo-Kronecker Functional Decision Diagrams) for a graph-based synthesis and is based on the number of nodes as the criterion of DD size. For a property of OPKFDD that is able to select one of different decomposition types for each node, OPKFDD is variable in its size by the decomposition types selection of each node and input variable order. This paper proposes a method for generating OPKFDD efficiently from the current BDD(Binary Decision Diagram) Data structure and an algorithm for minimizing one. In the multiple output functions, the relations of each function affect the number of nodes of OPKFDD. Therefore this paper proposes a method to decide the input variable order considering the above cases. Experimental results of comparing with the current representation methods and the reordering methods for deciding input variable order are shown.

• 한국통신학회논문지
• /
• 제30권5C호
• /
• pp.382-390
• /
• 2005

허프만 코드는 영상이나 비디오 전송뿐만 아니라 여러 분야에서 광범위하게 사용되고 있는 데이터 압축 알고리즘으로서, 실시간 데이터의 양이 증가함에 따라 효율적인 디코딩 알고리즘에 관한 많은 연구가 진행되고 있다. 본 논문에서는 호프만 디코딩을 위해 균형 트리를 형성하여 효율적인 이진 검색을 수행하는 구조를 제안하고 타 구조와의 성능을 비교하였다. 제안하는 구조는 길이가 다른 코드워드 간의 크기 비교를 가능하게 하는 정의를 사용하여 비어있는 내부 노드를 포함하지 않는 완전 균형 트리를 구성하므로, 디코딩 테이블을 위해 필요로 하는 메모리의 크기에 있어 매우 우수한 구조이다. 실제 영상 데이터를 사용하여 실험한 결과, 256개의 심볼 set에 대해 제안하는 구조는 매우 적은 수의 테이블 엔트리를 요구하며, 디코딩 성능은 최소 1번, 최대 5번, 평균 2.41번의 메모리 접근을 소요함을 보았다.