• Journal of KIISE:Software and Applications
• /
• v.34 no.1
• /
• pp.11-17
• /
• 2007

One of the popular methods used for pattern classification is the MBR (Memory-Based Reasoning) algorithm. Since it simply computes distances between a test pattern and training patterns or hyperplanes stored in memory, and then assigns the class of the nearest training pattern, it cannot explain how the classification result is obtained. In order to overcome this problem, we propose an incremental teaming algorithm based on RPA (Recursive Partition Averaging) to extract IF-THEN rules that describe regularities inherent in training patterns. But rules generated by RPA eventually show an overfitting phenomenon, because they depend too strongly on the details of given training patterns. Also RPA produces more number of rules than necessary, due to over-partitioning of the pattern space. Consequently, we present the IREA (Incremental Rule Extraction Algorithm) that overcomes overfitting problem by removing useless conditions from rules and reduces the number of rules at the same time. We verify the performance of proposed algorithm using benchmark data sets from UCI Machine Learning Repository.

• Radiation Oncology Journal
• /
• v.27 no.3
• /
• pp.140-144
• /
• 2009

Purpose: This study was performed to retrospectively analyze patient survival by weighting according to the primary tumor oncotype in 160 patients with brain metastasis and who underwent whole brain radiotherapy. Materials and Methods: A total of 160 metastatic brain cancer patients who were treated with whole brain radiotherapy of 30 Gy between 2002 and 2008 were retrospectively analyzed. The primary tumor oncotype of 20 patients was breast cancer, and that of 103 patients was lung cancer. Except for 18 patients with leptomeningeal seeding, a total of 142 patients were analyzed according to the prognostic factors and the Recursive Partitioning Analysis (RPA) class. Weighted Partitioning Analysis (WPA), with the weighting being done according to the primary tumor oncotype, was performed and the results were correlated with survival and then compared with the RPA Class. Results: The median survival of the patients in RPA Class I (8 patients) was 20.0 months, that for Class II (76 patients) was 10.0 months and that for Class III (58 patients) was 3.0 months (p<0.003). The median survival of patients in WPA Class I (3 patients) was 36 months, that for the patients in Class II (9 patients) was 23.7 months, that for the patients in Class III (70 patients) was 10.9 months and that for the patients in Class IV (60 patients) was 8.6 months (p<0.001). The WPA Class might have more accuracy in assessing survival, and it may be superior to the RPA Class for assessing survival. Conclusion: A new prognostic index, the WPA Class, has more prognostic value than the RPA Class for the treatment of patients with metastatic brain cancer. This WPA Class may be useful to guide the appropriate treatment of metastatic brain lesions.

• The KIPS Transactions:PartB
• /
• v.13B no.2 s.105
• /
• pp.127-132
• /
• 2006

K-NN (k-Nearest Neighbors), which is a well-known instance-based learning algorithm, simply stores entire training patterns in memory, and uses a distance function to classify a test pattern. K-NN is proven to show satisfactory performance, but it is notorious formemory usage and lengthy computation. Various studies have been found in the literature in order to minimize memory usage and computation time, and NGE (Nested Generalized Exemplar) theory is one of them. In this paper, we propose RPA (Recursive Partition Averaging) and IRPA (Incremental RPA) which is an incremental version of RPA. RPA partitions the entire pattern space recursively, and generates representatives from each partition. Also, due to the fact that RPA is prone to produce excessive number of partitions as the number of features in a pattern increases, we present IRPA which reduces the number of representative patterns by processing the training set in an incremental manner. Our proposed methods have been successfully shown to exhibit comparable performance to k-NN with a lot less number of patterns and better result than EACH system which implements the NGE theory.

• Communications for Statistical Applications and Methods
• /
• v.19 no.2
• /
• pp.293-301
• /
• 2012

Quantile regression proposed by Koenker and Bassett (1978) is a statistical technique that estimates conditional quantiles. The advantage of using quantile regression is the robustness in response to large outliers compared to ordinary least squares(OLS) regression. A regression tree approach has been applied to OLS problems to fit flexible models. Loh (2002) proposed the GUIDE algorithm that has a negligible selection bias and relatively low computational cost. Quantile regression can be regarded as an analogue of OLS, therefore it can also be applied to GUIDE regression tree method. Chaudhuri and Loh (2002) proposed a nonparametric quantile regression method that blends key features of piecewise polynomial quantile regression and tree-structured regression based on adaptive recursive partitioning. Lee and Lee (2006) investigated wage determinants in the Korean labor market using the Korean Labor and Income Panel Study(KLIPS). Following Lee and Lee, we fit three kinds of quantile regression tree models to KLIPS data with respect to the quantiles, 0.05, 0.2, 0.5, 0.8, and 0.95. Among the three models, multiple linear piecewise quantile regression model forms the shortest tree structure, while the piecewise constant quantile regression model has a deeper tree structure with more terminal nodes in general. Age, gender, marriage status, and education seem to be the determinants of the wage level throughout the quantiles; in addition, education experience appears as the important determinant of the wage level in the highly paid group.

• ETRI Journal
• /
• v.35 no.2
• /
• pp.270-280
• /
• 2013

A fast intra-prediction method is proposed for High Efficiency Video Coding (HEVC) using a fast intra-mode decision and fast coding unit (CU) size decision. HEVC supports very sophisticated intra modes and a recursive quadtree-based CU structure. To provide a high coding efficiency, the mode and CU size are selected in a rate-distortion optimized manner. This causes a high computational complexity in the encoder, and, for practical applications, the complexity should be significantly reduced. In this paper, among the many predefined modes, the intra-prediction mode is chosen without rate-distortion optimization processes, instead using the difference between the minimum and second minimum of the rate-distortion cost estimation based on the Hadamard transform. The experiment results show that the proposed method achieves a 49.04% reduction in the intra-prediction time and a 32.74% reduction in the total encoding time with a nearly similar coding performance to that of HEVC test model 2.1.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.40 no.12
• /
• pp.1195-1202
• /
• 1991

This paper describes a newly developed method of production simulation by using the Mixture of Cumulant Approximation (MOCA). In this method, the load is modelled as random variable (r.v.) which can be interpreted in terms of partitioning the load into various categories. We can consider the load shape of multi-modal characteristics. The number of load category and demarcation points of each load category are calculated automatically by using interpolation and least square method. Each generating unit of a supply system is modelled as r.v. of unit outage capacity according to the number of unit outage subset. Since the computation burden of each subset's moments increases exponentially as units are convolved to the system, we further derive the specific recursive formulae. In simulating the energy limited units, hydro unit simulation is performed using Energy Invariance Property and the simulation of pumped storage unit is modelled as compulsory and economic operations. The proposed MOCA method is applide to the test systems and the results are compared with those of cumulant and Booth Baleriaux method. It is verified that the MOCA method is considerably reliable and stable both pathological and well behaved system.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.104-123
• /
• 2022

Style transfer based on neural network provides very high quality results by reflecting the high level structural characteristics of images, and thereby has recently attracted great attention. This paper deals with the problem of resolution limitation due to GPU memory in performing such neural style transfer. We can expect that the gradient operation for style transfer based on partial image, with the aid of the fixed size of receptive field, can produce the same result as the gradient operation using the entire image. Based on this idea, each component of the style transfer loss function is analyzed in this paper to obtain the necessary conditions for partitioning and padding, and to identify, among the information required for gradient calculation, the one that depends on the entire input. By structuring such information for using it as auxiliary constant input for partition-based gradient calculation, this paper develops a recursive algorithm for super high-resolution image style transfer. Since the proposed method performs style transfer by partitioning input image into the size that a GPU can handle, it can perform style transfer without the limit of the input image resolution accompanied by the GPU memory size. With the aid of such super high-resolution support, the proposed method can provide a unique style characteristics of detailed area which can only be appreciated in super high-resolution style transfer.

• Journal of Broadcast Engineering
• /
• v.19 no.5
• /
• pp.627-630
• /
• 2014

The latest video coding standard, high efficiency video coding (HEVC) achieves high coding efficiency by employing a quadtree-based coding unit (CU) block partitioning structure which allows recursive splitting into four equally sized blocks. At each depth level, each CU is partitioned into variable sized blocks of prediction units (PUs). However, the determination of the best CU partition for each coding tree unit (CTU) and the best PU mode for each CU causes a dramatic increase in computational complexity. To reduce such computational complexity, we propose a fast PU decision algorithm that early terminates PU search. The proposed method skips the computation of R-D cost for certain PU modes in the current CU based on the best mode and the rate-distortion (RD) cost of the upper depth CU. Experimental results show that the proposed method reduces the computational complexity of HM12.0 to 18.1% with only 0.2% increases in BD-rate.

• Journal of Broadcast Engineering
• /
• v.22 no.5
• /
• pp.541-547
• /
• 2017

The Joint Exploration Model (JEM), which is a reference SW codec of the Joint Video Exploration Team (JVET) exploring the future video standard technology, provides a recursive Quadtree plus Binary Tree (QTBT) block structure. QTBT can achieve enhanced coding efficiency by adding new block structures at the expense of largely increased computational complexity. In this paper, we propose a fast decision algorithm of QTBT block partitioning depth that uses the rate-distortion (RD) cost of the upper and current depth to reduce the complexity of the JEM encoder. Experimental results showed that the computational complexity of JEM 5.0 can be reduced up to 21.6% and 11.0% with BD-rate increase of 0.7% and 1.2% in AI (All Intra) and RA (Random Access), respectively.

• Journal of Korean Neurosurgical Society
• /
• v.44 no.6
• /
• pp.358-363
• /
• 2008

Objective : This study was performed to assess the efficacy of GKS in patients with ten or more brain metastases. Methods : From Aug 2002 to Dec 2007, twenty-six patients (13 men and 13 women) with ten or more cerebral metastatic lesions underwent GKS. The mean age was 55 years (32-80). All patients had Karnofsky performance status (KPS) score of 70 or better. According to recursive partitioning analysis (RPA) classification, 3 patients belonged to class I and 23 to class II. The location of primary tumor was lung (21), breast (3) and unknown (2). The mean number of the lesions per patient was 16.6 (10-37). The mean cumulated volume was 10.9 cc (1.0-42.2). The median marginal dose was 15 Gy (9-23). Overall survival and the prognostic factors for the survival were retrospectively analyzed by using Kaplan Meier method and univariate analysis. Results : Overall median survival from GKS was 34 weeks (8-199). Local control was possible for 79.5% of the lesions and control of all the lesions was possible in at least 14 patients (53.8%) until 6 months after GKS. New lesions appeared in 7 (26.9%) patients during the same period. At the last follow-up, 18 patients died; 6 (33.3%) from systemic causes, 10 (55.6%) from neurological causes, and 2 (11.1 %) from unknown causes. Synchronous onset in non-small cell lung cancer (p=0.007), high KPS score (${\geq}80$, p=0.029), and controlled primary disease (p=0.020) were favorable prognostic factors in univariate analysis. Conclusion : In carefully selected patients, GKS may be a treatment option for ten or more brain metastases.