• Title/Summary/Keyword: Gradients

Search Result 1,176, Processing Time 0.032 seconds

Robustness of Differentiable Neural Computer Using Limited Retention Vector-based Memory Deallocation in Language Model

  • Lee, Donghyun;Park, Hosung;Seo, Soonshin;Son, Hyunsoo;Kim, Gyujin;Kim, Ji-Hwan
    • KSII Transactions on Internet and Information Systems (TIIS)
    • /
    • v.15 no.3
    • /
    • pp.837-852
    • /
    • 2021
  • Recurrent neural network (RNN) architectures have been used for language modeling (LM) tasks that require learning long-range word or character sequences. However, the RNN architecture is still suffered from unstable gradients on long-range sequences. To address the issue of long-range sequences, an attention mechanism has been used, showing state-of-the-art (SOTA) performance in all LM tasks. A differentiable neural computer (DNC) is a deep learning architecture using an attention mechanism. The DNC architecture is a neural network augmented with a content-addressable external memory. However, in the write operation, some information unrelated to the input word remains in memory. Moreover, DNCs have been found to perform poorly with low numbers of weight parameters. Therefore, we propose a robust memory deallocation method using a limited retention vector. The limited retention vector determines whether the network increases or decreases its usage of information in external memory according to a threshold. We experimentally evaluate the robustness of a DNC implementing the proposed approach according to the size of the controller and external memory on the enwik8 LM task. When we decreased the number of weight parameters by 32.47%, the proposed DNC showed a low bits-per-character (BPC) degradation of 4.30%, demonstrating the effectiveness of our approach in language modeling tasks.

Validation of Gamma Knife Perfexion Dose Profile Distribution by a Modified Variable Ellipsoid Modeling Technique

  • Hur, Beong Ik;Jin, Seong Jin;Kim, Gyeong Rip;Kwak, Jong Hyeok;Kim, Young Ha;Lee, Sang Weon;Sung, Soon Ki
    • Journal of Korean Neurosurgical Society
    • /
    • v.64 no.1
    • /
    • pp.13-22
    • /
    • 2021
  • Objective : High precision and accuracy are expected in gamma knife radiosurgery treatment. Because of the requirement of clinically applying complex radiation and dose gradients together with a rapid radiation decline, a dedicated quality assurance program is required to maintain the radiation dosimetry and geometric accuracy and to reduce all associated risk factors. This study investigates the validity of Leksell Gamma plan (LGP)10.1.1 system of 5th generation Gamma Knife Perfexion as modified variable ellipsoid modeling technique (VEMT) method. Methods : To verify LGP10.1.1 system, we compare the treatment plan program system of the Gamma Knife Perfexion, that is, the LGP, with the calculated value of the proposed modified VEMT program. To verify a modified VEMT method, we compare the distributions of the dose of Gamma Knife Perfexion measured by Gafchromic EBT3 and EBT-XD films. For verification, the center of an 80 mm radius solid water phantom is placed in the center of all sectors positioned at 16 mm, 4 mm and 8 mm; that is, the dose distribution is similar to the method used in the x, y, and z directions by the VEMT. The dose distribution in the axial direction is compared and analyzed based on Full-Width-of-Half-Maximum (FWHM) evaluation. Results : The dose profile distribution was evaluated by FWHM, and it showed an average difference of 0.104 mm for the LGP value and 0.130 mm for the EBT-XD film. Conclusion : The modified VEMT yielded consistent results in the two processes. The use of the modified VEMT as a verification tool can enable the system to stably test and operate the Gamma Knife Perfexion treatment planning system.

Implementation of Spatial Downscaling Method Based on Gradient and Inverse Distance Squared (GIDS) for High-Resolution Numerical Weather Prediction Data (고해상도 수치예측자료 생산을 위한 경도-역거리 제곱법(GIDS) 기반의 공간 규모 상세화 기법 활용)

  • Yang, Ah-Ryeon;Oh, Su-Bin;Kim, Joowan;Lee, Seung-Woo;Kim, Chun-Ji;Park, Soohyun
    • Atmosphere
    • /
    • v.31 no.2
    • /
    • pp.185-198
    • /
    • 2021
  • In this study, we examined a spatial downscaling method based on Gradient and Inverse Distance Squared (GIDS) weighting to produce high-resolution grid data from a numerical weather prediction model over Korean Peninsula with complex terrain. The GIDS is a simple and effective geostatistical downscaling method using horizontal distance gradients and an elevation. The predicted meteorological variables (e.g., temperature and 3-hr accumulated rainfall amount) from the Limited-area ENsemble prediction System (LENS; horizontal grid spacing of 3 km) are used for the GIDS to produce a higher horizontal resolution (1.5 km) data set. The obtained results were compared to those from the bilinear interpolation. The GIDS effectively produced high-resolution gridded data for temperature with the continuous spatial distribution and high dependence on topography. The results showed a better agreement with the observation by increasing a searching radius from 10 to 30 km. However, the GIDS showed relatively lower performance for the precipitation variable. Although the GIDS has a significant efficiency in producing a higher resolution gridded temperature data, it requires further study to be applied for rainfall events.

Elevational distribution ranges of vascular plant species in the Baekdudaegan mountain range, South Korea

  • Shin, Sookyung;Kim, Jung-Hyun;Dang, Ji-Hee;Seo, In-Soon;Lee, Byoung Yoon
    • Journal of Ecology and Environment
    • /
    • v.45 no.1
    • /
    • pp.62-71
    • /
    • 2021
  • The climate is changing rapidly, and this may pose a major threat to global biodiversity. One of the most distinctive consequences of climate change is the poleward and/or upward shift of species distribution ranges associated with increasing temperatures, resulting in a change of species composition and community structure in the forest ecosystems. The Baekdudaegan mountain range connects most forests from the lowland to the subalpine zone in South Korea and is therefore recognized as one of the most important biodiversity hotspots. This study was conducted to understand the distribution range of vascular plants along elevational gradients through field surveys in the six national parks of the Baekdudaegan mountain range. We identified the upper and lower distribution limits of a total of 873 taxa of vascular plants with 117 families, 418 genera, 793 species, 14 subspecies, 62 varieties, two forms, and two hybrids. A total of 12 conifers were recorded along the elevational gradient. The distribution ranges of Abies koreana, Picea jezoensis, Pinus pumila, and Thuja koraiensis were limited to over 1000 m above sea level. We also identified 21 broad-leaved trees in the subalpine zone. A total of 45 Korean endemic plant species were observed, and of these, 15 taxa (including Aconitum chiisanense and Hanabusaya asiatica) showed a narrow distribution range in the subalpine zone. Our study provides valuable information on the current elevational distribution ranges of vascular plants in the six national parks of South Korea, which could serve as a baseline for vertical shifts under future climate change.

Performance Comparison of VRS and FKP Network RTK User According to Baseline Length (기선 거리에 따른 VRS와 FKP 방식의 Network RTK 사용자 성능 비교)

  • Lim, Cheolsoon;Park, Byungwoon
    • Journal of Advanced Navigation Technology
    • /
    • v.24 no.6
    • /
    • pp.540-548
    • /
    • 2020
  • In this paper, the performances of virtual reference station (VRS) and flächen korrektur parameter (FKP) based Network real time kinematics (RTK) according to baseline length were compared and analyzed. We applied the VRS and FKP corrections for each baseline length obtained from National Geographic Information Institute Network RTK services to an FKP-supported commercial receiver and analyzed the RTK results in the range and position domains. In the case of VRS, RTK performance was degraded due to the spatial error, which increase in proportion of the baseline length. On the other hand, FKP compensates for spatial errors by using the gradients of dispersive and non-dispersive errors, so it showed stable RTK performance compared to VRS even if the baseline length increases up to 130 km. However, in the case of long baseline of 150 km or more, integer ambiguities were incorrectly fixed due to the decrease in the performance of the FKP corrections.

Linking Leaf Functional Traits with Plant Resource Utilization Strategy in an Evergreen Scrub Species Rhododendron caucasicum Pall. along Longitudinal Gradient in Georgia (The South Caucasus)

  • Ekhvaia, Jana;Bakhia, Arsena;Asanidze, Zezva;Beltadze, Tornike;Abdaladze, Otar
    • Journal of Forest and Environmental Science
    • /
    • v.38 no.2
    • /
    • pp.110-121
    • /
    • 2022
  • Leaf functional traits widely have been used to understand the environmental controls of resource utilization strategy of plants along the environmental gradients. By using key leaf functional traits, we quantified the relationships between leaf traits and local climate throughout the distributional range of Rhododendron caucasicum Pall. in eastern and western Georgian mountains (the South Caucasus). Our results revealed, that all traits showed high levels of intraspecific variability across study locations and confirmed a strong phenotypic differentiation of leaf functional variation along the east-west longitudinal gradient in response to the local climate; out of the explored climatic variables, the moisture factors related to precipitation and number of precipitation and dry days for winter and growth seasons were more strongly related to leaf trait variation than the elevation and air temperature. Among studied leaf traits, the leaf specific area (SLA) showed the highest level of variability indicating the different resource utilization strategies of eastern and western-central Rh. caucasicum individuals. High SLA leaves for western-central Caucasian individuals work in relatively resource-rich environments (more humid in terms of precipitation amount and the number of precipitation days in winter) and could be explained by preferential allocation to photosynthesis and growth, while eastern Caucasian samples work in resource-poor environments (less humid in terms of precipitation amount and the number of precipitation days in winter) and the retention of captured resources is a higher priority appearing in a low SLA leaves. However, more evidence from a broader study of the species throughout its distribution range by including additional environmental factors and molecular markers are needed for firmer conclusions of intraspecific variability of Rh. caucasicum.

High Temperature Drying of Pitch Pine Lumber (리기다소나무 고온건조)

  • Yeo, Hwan-Myeong;Shim, Sang-Ro
    • Journal of the Korean Wood Science and Technology
    • /
    • v.32 no.4
    • /
    • pp.46-51
    • /
    • 2004
  • The objective of this study was to evaluate the applicability of high temperature drying to pitch pine (Pinus rigida) lumber, especially intended for use as lamina of structural glued laminated timber (glulam), to enhance the efficient utilization and provide added-value to that species. The high temperature schedule and drying procedures utilized were shown to be reasonable for drying glulam lamina due to the occurrence of small moisture gradients, minimal residual drying stress, and low warpage. Through preliminary tests, it was confirmed that residual resin at lamina surfaces did not adversely affect the gluing process. However, quantitative analysis of resin is required for developing a method to constrain the occurrence of pitch trouble with respect to decreasing long-term adhesive and finish durability of glulam in use after manufacturing. The final moisture content of high temperature dried lamina was much lower than target moisture content and discoloration was more severe than anticipated. In a further study, it will be necessary to determine the optimal drying conditions, such as temperature, humidity, drying time, and top load restraint level, which could best control discoloration, warpage, and moisture content of the lamina, while minimizing fuel expense.

A quasi 3D solution for thermodynamic response of FG sandwich plates lying on variable elastic foundation with arbitrary boundary conditions

  • Bouiadjra, Rabbab Bachir;Mahmoudi, Abdelkader;Sekkal, Mohamed;Benyoucef, Samir;Selim, Mahmoud M.;Tounsi, Abdelouahed;Hussain, Muzamal
    • Steel and Composite Structures
    • /
    • v.41 no.6
    • /
    • pp.873-886
    • /
    • 2021
  • In this paper, an analytical solution for thermodynamic response of functionally graded (FG) sandwich plates resting on variable elastic foundation is performed by using a quasi 3D shear deformation plate theory. The displacement field used in the present study contains undetermined integral terms and involves only four unknown functions with including stretching effect. The FG sandwich plate is considered to be subject to a time harmonic sinusoidal temperature field across its thickness with any combined boundary conditions. Equations of motion are derived from Hamilton's principle. The numerical results are compared with the existing results of quasi-3D shear deformation theories and an excellent agreement is observed. Several numerical examples for fundamental frequency, deflection, stress and variable elastic foundation parameter's analysis of FG sandwich plates are presented and discussed considering different material gradients, layer thickness ratios, thickness-to-length ratios and boundary conditions. The results of the present study reveal that the nature of the elastic foundation, the boundary conditions and the thermodynamic loading affect the response of the FG plate especially in the case of a thick plate.

Properties of CoGe thin film-based galvanic cells and their applications for IoT sensor networks (CoGe 박막 기반 galvanic cell의 특성 및 IoT 센서 네트워크에 대한 적용)

  • Jeon, Buil;Han, Dongsoo;Yoon, Giwan
    • Journal of the Korea Institute of Information and Communication Engineering
    • /
    • v.26 no.9
    • /
    • pp.1347-1356
    • /
    • 2022
  • In this paper, we investigate the properties of CoGe thin film-based galvanic cells as a function of their dimension (cell length, width, etc.) and show their application as sensors to Arduino-based IoT sensor networks to detect water contact. Because these CoGe thin film-based galvanic cells do not require mechanical strains or temperature gradients unlike piezoelectric and thermoelectric energy harvesters, we think that these thin film-based galvanic cells are more suitable for self-powered sensor networks demanding sustainable and robust energy harvesters. In the past, a sputter-deposited CoGe thin film has not been intensively investigated for energy harvesting appilcations. Thus, in this study, we perform a feasibility study of galvanic cells composed of a sputter-deposited CoGe thin film to see if they can be applied as potential self-powered sensors. We believe that this paper will be of great help in developing even more enhanced sensor networks.

Environmental test campaign of a 6U CubeSat Test Platform equipped with an ambipolar plasma thruster

  • Stesina, Fabrizio;Corpino, Sabrina;Borras, Eduard Bosch;Amo, Jose Gonzalez Del;Pavarin, Daniele;Bellomo, Nicolas;Trezzolani, Fabio
    • Advances in aircraft and spacecraft science
    • /
    • v.9 no.3
    • /
    • pp.195-215
    • /
    • 2022
  • The increasing interest in CubeSat platforms ant their capability of enlarging the frontier of possible missions impose technology improvements. Miniaturized electrical propulsion (EP) systems enable new mission for multi-unit CubeSats (6U+). While electric propulsion systems have achieved important level of knowledge at equipment level, the investigation of the mutual impact between EP system and CubeSat technology at system level can provide a decisive improvement for both the technologies. The interaction between CubeSat and EP system should be assessed in terms of electromagnetic emissions (both radiated and conducted), thermal gradients, high electrical power management, surface chemical deposition, and quick and reliable data exchanges. This paper shows how a versatile CubeSat Test Platform (CTP), together with standardized procedures and specialized facilities enable the acquisition fundamental and unprecedented information. Measurements can be taken both by specific ground support equipment placed inside the vacuum facility and by dedicated sensors and subsystems installed on the CTP, providing a completely new set of data never obtained before. CTP is constituted of a 6U primary structure hosting the EP system, representative CubeSat avionics and batteries. For the first test campaign, CTP hosts the ambipolar plasma propulsion system, called Regulus and developed by T4I. After the integration and the functional test in laboratory environment, CTP + Regulus performed a Test campaign in relevant environment in the vacuum chamber at CISAS, University of Padua. This paper is focused on the test campaign description and the main results achieved at different power levels for different duration of the firings.