• ETRI Journal
• /
• 제41권3호
• /
• pp.371-382
• /
• 2019

Machine reading comprehension is the task of understanding a given context and finding the correct response in that context. A simple recurrent unit (SRU) is a model that solves the vanishing gradient problem in a recurrent neural network (RNN) using a neural gate, such as a gated recurrent unit (GRU) and long short-term memory (LSTM); moreover, it removes the previous hidden state from the input gate to improve the speed compared to GRU and LSTM. A self-matching network, used in R-Net, can have a similar effect to coreference resolution because the self-matching network can obtain context information of a similar meaning by calculating the attention weight for its own RNN sequence. In this paper, we construct a dataset for Korean machine reading comprehension and propose an $S^2-Net$ model that adds a self-matching layer to an encoder RNN using multilayer SRU. The experimental results show that the proposed $S^2-Net$ model has performance of single 68.82% EM and 81.25% F1, and ensemble 70.81% EM, 82.48% F1 in the Korean machine reading comprehension test dataset, and has single 71.30% EM and 80.37% F1 and ensemble 73.29% EM and 81.54% F1 performance in the SQuAD dev dataset.

• Journal of Forest and Environmental Science
• /
• 제35권2호
• /
• pp.90-101
• /
• 2019

Biodiversity refers to the total number and variation among species of flora and fauna of an area. Due to tremendous biotic especially anthropogenic pressure these natural resources are being vanishing. In present study genetic diversity among accessions of Thamnocalamus spathiflorus was evaluated. A total of 51 vegetative characters and 42 primers (10-mer) were screened. Out of 42 screened primers, 28 polymorphic primers were selected for further analysis. A total of 263 bands were recorded as polymorphic whereas 48 bands were monomorphic. The resolving power (Rp) of 28 Randomly Amplified Polymorphic DNA (RAPD) primers ranged from 4.6 (OPE08) to 17.6 (OPA11). The polymorphic information content (PIC) value ranged from 0.21 (OPAH09) to 0.44 (OPG02). The result revealed high degree of genetic relatedness (56 to 80%). Cluster analysis revealed two major clusters both for morphology as well as RAPD. Unlike morphological characterization, the accession (D5) from Bahli, Rampur, Shimla (H.P.) was clustered separately from the others in RAPD cluster analysis. Accessions with closed locality grouped together through RAPD marker system however analogy was recorded for morphological traits. The study conducted reflects the utility of RAPD technique for species identification and phylogenetic studies in bamboo for conducting bamboo breeding program.

• 영미문화
• /
• 제10권2호
• /
• pp.99-125
• /
• 2010

This paper explores the political significance of a literary work, the hidden side beneath the ideology of founding America in Lydia Maria Child's Hobomok which reconstructs the history of the colonial period. The ideological strategy of founding America on racial discrimination is given a repeated representation in 19th-century American novels. Most works shed a negative light on Native Americans, whereas Hobomok stands out by presenting a positive picture of a miscegenation between a Native American man and a white woman, the acculturation of a half Indian into the white society. Furthermore, Child undoes distorted stereotypes about native Americans, exposing the Puritans' intolerant and exclusive attitudes and criticizing men who forced women to be obedient for the cause of nation and religion. However, Child also shows that she could not be free from the ideology of founding America which insisted on the superiority of the white's racial identity and excluded the Native Americans as beings who were destined to vanish gradually but eventually. Although Hobomok revises stereotypical representation of Native Americans as the other, it also serves for a political purpose, showing a politically inseparable relationship between literary works and the ideology of founding America.

• Kyungpook Mathematical Journal
• /
• 제60권4호
• /
• pp.683-710
• /
• 2020

Let ℕ be the set of nonnegative integers and 𝕬 be a 2-torsion free triangular algebra over a commutative ring ℛ. In the present paper, under some lenient assumptions on 𝕬, it is proved that if Δ = {𝛿_n}_n∈ℕ is a sequence of ℛ-linear mappings 𝛿_n : 𝕬 → 𝕬 satisfying ${\delta}_n([[x,\;y],\;z])\;=\;\displaystyle\sum_{i+j+k=n}\;[[{\delta}_i(x),\;{\delta}_j(y)],\;{\delta}_k(z)]$ for all x, y, z ∈ 𝕬 with xy = 0 (resp. xy = p, where p is a nontrivial idempotent of 𝕬), then for each n ∈ ℕ, 𝛿_n = d_n + 𝜏_n; where d_n : 𝕬 → 𝕬 is ℛ-linear mapping satisfying $d_n(xy)\;=\;\displaystyle\sum_{i+j=n}\;d_i(x)d_j(y)$ for all x, y ∈ 𝕬, i.e. 𝒟 = {d_n}_n∈ℕ is a higher derivation on 𝕬 and 𝜏_n : 𝕬 → Z(𝕬) (where Z(𝕬) is the center of 𝕬) is an ℛ-linear map vanishing at every second commutator [[x, y], z] with xy = 0 (resp. xy = p).

• International Journal of Computer Science & Network Security
• /
• 제21권11호
• /
• pp.17-22
• /
• 2021

Recently, image super-resolution techniques used in convolutional neural networks (CNN) have led to remarkable performance in the research area of digital image processing applications and computer vision tasks. Convolutional layers stacked on top of each other can design a more complex network architecture, but they also use more memory in terms of the number of parameters and introduce the vanishing gradient problem during training. Furthermore, earlier approaches of single image super-resolution used interpolation technique as a pre-processing stage to upscale the low-resolution image into HR image. The design of these approaches is simple, but not effective and insert the newer unwanted pixels (noises) in the reconstructed HR image. In this paper, authors are propose a novel single image super-resolution architecture based on synchronized depthwise separable convolution with Dense Skip Connection Block (DSCB). In addition, unlike existing SR methods that only rely on single path, but our proposed method used the synchronizes path for generating the SISR image. Extensive quantitative and qualitative experiments show that our method (SDCN) achieves promising improvements than other state-of-the-art methods.

• 대한수학회보
• /
• 제59권5호
• /
• pp.1145-1166
• /
• 2022

In this paper, for a transcendental meromorphic function f and α ∈ ℂ, we have exhaustively studied the nature and form of solutions of a new type of non-linear differential equation of the following form which has never been investigated earlier: $$f^n+{\alpha}f^{n-2}f^{\prime}+P_d(z,f)={\sum\limits_{i=1}^{k}}{p_i(z)e^{{\alpha}_i(z)},$$ where P_d(z, f) is a differential polynomial of f, p_i's and α_i's are non-vanishing rational functions and non-constant polynomials, respectively. When α = 0, we have pointed out a major lacuna in a recent result of Xue [17] and rectifying the result, presented the corrected form of the same equation at a large extent. In addition, our main result is also an improvement of a recent result of Chen-Lian [2] by rectifying a gap in the proof of the theorem of the same paper. The case α ≠ 0 has also been manipulated to determine the form of the solutions. We also illustrate a handful number of examples for showing the accuracy of our results.

• International journal of advanced smart convergence
• /
• 제12권4호
• /
• pp.142-146
• /
• 2023

We present a method for generating 3D structures and rendering objects by combining VAE (Variational Autoencoder) and GAN (Generative Adversarial Network). This approach focuses on generating and rendering 3D models with improved quality using residual learning as the learning method for the encoder. We deep stack the encoder layers to accurately reflect the features of the image and apply residual blocks to solve the problems of deep layers to improve the encoder performance. This solves the problems of gradient vanishing and exploding, which are problems when constructing a deep neural network, and creates a 3D model of improved quality. To accurately extract image features, we construct deep layers of the encoder model and apply the residual function to learning to model with more detailed information. The generated model has more detailed voxels for more accurate representation, is rendered by adding materials and lighting, and is finally converted into a mesh model. 3D models have excellent visual quality and accuracy, making them useful in various fields such as virtual reality, game development, and metaverse.

• 대한수학회논문집
• /
• 제39권1호
• /
• pp.187-199
• /
• 2024

In the present paper, following the pullback approach to Finsler geometry, we study intrinsically the C^v-reducible and generalized C^v-reducible Finsler spaces. Precisely, we introduce a coordinate-free formulation of these manifolds. Then, we prove that a Finsler manifold is C^v-reducible if and only if it is C-reducible and satisfies the 𝕋-condition. We study the generalized C^v-reducible Finsler manifold with a scalar π-form 𝔸. We show that a Finsler manifold (M, L) is generalized C^v-reducible with 𝔸 if and only if it is C-reducible and 𝕋 = 𝔸. Moreover, we prove that a Landsberg generalized C^v-reducible Finsler manifold with a scalar π-form 𝔸 is Berwaldian. Finally, we consider a special C^v-reducible Finsler manifold and conclude that a Finsler manifold is a special C^v-reducible if and only if it is special semi-C-reducible with vanishing 𝕋-tensor.

• 정보처리학회논문지B
• /
• 제19B권4호
• /
• pp.209-220
• /
• 2012

본 논문에서는 보행자에 장착된 카메라로부터 입력된 자연영상에서의 구조적 패턴 변화에 강인한 적응적인 보도 영역 검출 기법을 제안한다. 제안하는 방법에서는 다양한 패턴을 가지는 보도 환경에서 안정적으로 보도 영역을 분할하기 위해 첫 번째 단계에서는 소실점에 기반하는 VRay를 이용한 방사형 영역 분할법을 통해 보도의 경계선을 검출하여 보도의 후보영역을 분리하며, 두 번째 단계에서는 분리된 후보영역 내에서의 시드 영역 확장법(SRG)을 개선한 적응적 다중 시드 영역 확장법(A-MSRG)를 통해 구조적 패턴이 반복되는 보도 영역을 실시간으로 검출하는 방법을 수행한다. 성능평가를 위해 제안된 방사형 영역 분할법과 A-MSRG와의 결합에 의한 영역 검출 결과의 효율성을 측정한다. 기존의 SRG, MSRG 방법과의 비교 수행을 통해 제안된 방법의 타당성을 입증하였다.

• International Journal of Fluid Machinery and Systems
• /
• 제3권2호
• /
• pp.113-121
• /
• 2010

LES(Large Eddy Simulation) with a cavitation model was performed to calculate an unsteady flow for a mixed flow pump with a closed type impeller. First, the comparison between the numerical and experimental results was done to evaluate a computational accuracy. Second, the torque acting on the blade was calculated by simulation to investigate how the cavitation caused the fluctuation of torque. The absolute pressure around the leading edge on the suction side of blade surface had positive impulsive peaks in both the numerical and experimental results. The simulation showed that those peaks were caused by the cavitaion which contracted and vanished around the leading edge. The absolute pressure was predicted by simulation with -10% error. The absolute pressure around the trailing edge on the suction side of blade surface had no impulsive peaks in both the numerical and experimental results, because the absolute pressure was 100 times higher than the saturated vapor pressure. The simulation results showed that the cavitation was generated around the throat, then contracted and finally vanished. The simulated pump had five throats and cavitation behaviors such as contraction and vanishing around five throats were different from each other. For instance, the cavitations around those five throats were not vanished at the same time. When the cavitation was contracted and finally vanished, the absolute pressure on the blade surface was increased. When the cavitation was contracted around the throat located on the pressure side of blade surface, the pressure became high on the pressure side of blade surface. It caused the 1.4 times higher impulsive peak in the torque than the averaged value. On the other hand, when the cavitation was contracted around the throat located on the suction side of blade surface, the pressure became high on the suction side of blade surface. It caused the 0.4 times lower impulsive peak in the torque than the averaged value. The cavitation around the throat caused the large fluctuation in torque acting on the blade.