• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.10
• /
• pp.2643-2657
• /
• 2023

Scene graphs are structured representations that can clearly convey objects and the relationships between them, but are often heavily biased due to the highly skewed, long-tailed relational labeling in the dataset. Indeed, the visual world itself and its descriptions are biased. Therefore, Unbiased Scene Graph Generation (USGG) prefers to train models to eliminate long-tail effects as much as possible, rather than altering the dataset directly. To this end, we propose Geometric and Semantic Improvement (GSI) for USGG to mitigate this issue. First, to fully exploit the feature information in the images, geometric dimension and semantic dimension enhancement modules are designed. The geometric module is designed from the perspective that the position information between neighboring object pairs will affect each other, which can improve the recall rate of the overall relationship in the dataset. The semantic module further processes the embedded word vector, which can enhance the acquisition of semantic information. Then, to improve the recall rate of the tail data, the Class Balanced Seesaw Loss (CBSLoss) is designed for the tail data. The recall rate of the prediction is improved by penalizing the body or tail relations that are judged incorrectly in the dataset. The experimental findings demonstrate that the GSI method performs better than mainstream models in terms of the mean Recall@K (mR@K) metric in three tasks. The long-tailed imbalance in the Visual Genome 150 (VG150) dataset is addressed better using the GSI method than by most of the existing methods.

• Journal of KIISE:Computer Systems and Theory
• /
• v.27 no.2
• /
• pp.169-176
• /
• 2000

This paper is considered with the problem of embedding complete binary trees into 3-dimensional meshes using dimension-ordered routing with primary concern of minimizing link congestion. The authors showed that a complete binary tree with $2^P-1$ nodes can be embedded into a 3-dimensional mesh with optimum size, $2^P$ nodes, if the link congestion is two[14], (More precisely, the link congestion of each dimension is two, two, and one if the dimension-ordered routing is used, and two, one, and one if the dimension-ordered routing is not imposed.) In this paper, we present a scheme to find an embedding of a complete binary tree into a 3-dimensional mesh of size no larger than 1.27 times the optimum with link congestion one while using dimension-ordered routing.

• Journal of Environmental Science International
• /
• v.11 no.9
• /
• pp.915-923
• /
• 2002

Recently, GIS(Geographic Information System) is used to extract various hydrological factors from DEM(Digital Elevation Model) in river basin. Therefore, this study aims at the determination of river fractal dimension using DEM. In this paper, the main-stream length in river basin was grid-analyzed for each scale(1/5,000, 1/25,000, 1/50,000) and each cell size(5m$\times$5m, l0m$\times$l0m, 20m$\times$20m, 30m$\times$30m, 40m$\times$40m, 50m$\times$50m, 60m$\times$60m, 70m$\times$70m, 80m$\times$80m, 90m$\times$90m, 100m$\times$l00m, 120m$\times$120m, 150m$\times$150m) using GIS. Also, fractal dimension was derived by analyzing correlation among main-stream lengths, scale, and cell size which were calculated here. The result of calculating fractal dimension for each cell size shows that the fractal dimension on the main-stream length is 1.028.

• Wind and Structures
• /
• v.31 no.4
• /
• pp.363-371
• /
• 2020

Proper understanding of offshore wind speed variability is of essential importance in practice, which provides useful information to a wide range of coastal and marine activities. In this paper, long-term wind speed data recorded at various offshore stations are analyzed in the framework of fractal dimension analysis. Fractal analysis is a well-established data analysis tool, which is particularly suitable to determine the complexity in time series from a quantitative point of view. The fractal dimension is estimated using the conventional box-counting method. The results suggest that the wind speed data are generally fractals, which are likely to exhibit a persistent nature. The mean fractal dimension varies from 1.31 at an offshore weather station to 1.43 at an urban station, which is mainly associated with surface roughness condition. Monthly variability of fractal dimension at offshore stations is well-defined, which often possess larger values during hotter months and lower values during winter. This is partly attributed to the effect of thermal instability. In addition, with an increase in measurement interval, the mean and minimum fractal dimension decrease, whereas the maximum and coefficient of variation increase in parallel.

• Journal of Fisheries and Marine Sciences Education
• /
• v.29 no.2
• /
• pp.447-452
• /
• 2017

A closely associated set of characteristics was analyzed using both classical and truss dimensions to discriminate sexual dimorphism between spawning season and non spawning season in grass puffer, Takifugu niphobles. In non-spawning season, $1{\times}10/Ls$ of classical dimension was significantly different between male and female (P<0.05). In spawning season, $1{\times}9/Ls$ and $1{\times}10/Ls$ of classical dimension, $3{\times}8/Ls$, $3{\times}9/Ls$, $3{\times}10/Ls$, $4{\times}8/Ls$, $4{\times}9/Ls$ and $8{\times}9/Ls$ of truss dimension and $7{\times}9/HL$ of head part dimension were significantly different between male and female (P<0.05). The results of this study may be useful as basic information of other fish species to compare the change of sexual dimorphism between spawning season and non spawning season.

• Journal of Chest Surgery
• /
• v.27 no.8
• /
• pp.656-663
• /
• 1994

Evaluation of heart function is of importance in assessing the results of valvular heart surgery. Information on volume and functional change of heart chamber can be obtained by cardiac catheterization and echocardiography. We studied 41 patients with mitral stenosis[MS] and 23 patients with mitral regurgitation[MR] using M-mode echocardiography before and after mitral valve replacement[MVR] at Pusan Paik Hospital. Preoperative cardiac catheterization was available in 56 cases, and the results were obtained as follows. 1. In patients with MS, preoperative average LV end-diastolic dimension[EDD] and end-systolic dimension[ESD] were remained within normal range, but postoperative EDD and ESD were significantly decreased[P<0.01]. The preoperative and postoperative LV ejection fraction[EF] were remained within the normal range and no significant change[P>0.05]. The preoperative left atrial dimension[LAD] was enlarged considerably above normal[P<0.01], but was significantly decreased after surgery[P<0.001]. The preoperative LV posterior wall thickness[PWTh] was within normal range, and no significant change after surgery[P>0.05]. 2. In patients with MR, preoperative average end-diastolic dimension[EDD] and end-systolic dimension[ESD] were significantly greater than normal[P<005], but postoperative EDD and ESD were significantly decreased[P<0.01]. The preoperative LV ejection fraction[EF] and fractional shortening[FS] were within normal range, and no significant change after surgery[P>0.05].The preoperative left atrial dimension[LAD] was enlarged considerably above normal [P<0.01], but was significantly decreased after surgery[P<0.001].The preoperative LV posterior wall thickness[PWTh] was within normal range, and no significant change after surgery[P>0.05].

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.1
• /
• pp.1-8
• /
• 2004

According to the development of the Internet and the pervasion of Data Base, it is not easy to search for necessary information from the huge amounts of data. In order to do efficient analysis of a large amounts of data, this paper proposes a method for pattern classification based on the effective strategy for dimension reduction for narrowing down the whole data to what users wants to search for. To analyze data effectively, Radial Basis Function Networks based on VC-dimension of Support Vector Machine, a model of statistical teaming, is proposed in this paper. The model of Radial Basis Function Networks currently used performed the preprocessing of Perceptron model whereas the model proposed in this paper, performing independent analysis on VD-dimension, classifies each datum putting precise labels on it. The comparison and estimation of various models by using Machine Learning Data shows that the model proposed in this paper proves to be more efficient than various sorts of algorithm previously used.

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.147-150
• /
• 2012

While LDA is a supervised dimension reduction method which finds projective directions to maximize separability between classes, the performance of LDA is severely degraded when the number of labeled data is small. Recently semi-supervised dimension reduction methods have been proposed which utilize abundant unlabeled data and overcome the shortage of labeled data. However, matrix computation usually used in statistical dimension reduction methods becomes hindrance to make the utilization of a large number of unlabeled data difficult, and moreover too much information from unlabeled data may not so helpful compared to the increase of its processing time. In order to solve these problems, we propose an ensemble approach for semi-supervised dimension reduction. Extensive experimental results in text classification demonstrates the effectiveness of the proposed method.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.7A
• /
• pp.623-631
• /
• 2010

In this paper, various of pilot-assisted channel estimation techniques for 3GPP LTE downlink are tested under multipath Rayleigh fading channel. At first, the conventional channel estimation techniques are applied with linear zero-forcing (ZF) equalizer, such as one dimensional least square (1-D LS) linear interpolation, two dimensional (2-D) wiener filter, the time and frequency dimension separate wiener filter and maximum likelihood estimator (MLE). Considering the practical implementation, we proposed two channel estimation techniques by combining time-dimension wiener filter and MLE in two manners, which showed a good tradeoff between system performance and complexity when comparing with conventional techniques. The nonlinear decision feedback equalizer (DFE) which can show a better performance than linear ZF equalizer is also implemented for mitigating inter-carrier interference (ICI) in our system. The complexity of these algorithms are calculated in terms of the number of complex multiplications (CMs) and the performances are evaluated by showing the bit error rate (BER).

• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.1988-1991
• /
• 2002

In this paper, we propose a method for realizing transfer functions containing variable parameter, by the state-space method. By using this method, variable transfer functions (VTF) can be often realized with a minimal dimension. In case that a minimal realization can not be obtained, the realization dimension can be fairly reduced.