• The Korean Journal of Applied Statistics
• /
• v.32 no.5
• /
• pp.693-702
• /
• 2019

Deep learning has gained popularity for the classification and prediction task. Neural network layers become deeper as more data becomes available. Saturation is the phenomenon that the gradient of an activation function gets closer to 0 and can happen when the value of weight is too big. Increased importance has been placed on the issue of saturation which limits the ability of weight to learn. To resolve this problem, Glorot and Bengio (Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, 249-256, 2010) claimed that efficient neural network training is possible when data flows variously between layers. They argued that variance over the output of each layer and variance over input of each layer are equal. They proposed a method of initialization that the variance of the output of each layer and the variance of the input should be the same. In this paper, we propose a new method of establishing initialization by adopting truncated normal distribution and truncated cauchy distribution. We decide where to truncate the distribution while adapting the initialization method by Glorot and Bengio (2010). Variances are made over output and input equal that are then accomplished by setting variances equal to the variance of truncated distribution. It manipulates the distribution so that the initial values of weights would not grow so large and with values that simultaneously get close to zero. To compare the performance of our proposed method with existing methods, we conducted experiments on MNIST and CIFAR-10 data using DNN and CNN. Our proposed method outperformed existing methods in terms of accuracy.

• Nuclear Medicine and Molecular Imaging
• /
• v.43 no.5
• /
• pp.451-458
• /
• 2009

Purpose: Optical molecular luminescence imaging is widely used for detection and imaging of bio-photons emitted by luminescent luciferase activation. The measured photons in this method provide the degree of molecular alteration or cell numbers with the advantage of high signal-to-noise ratio. To extract useful information from the measured results, the analysis based on a proper quantification method is necessary. In this research, we propose a quantification method presenting linear response of measured light signal to measurement time. Materials and Methods: We detected the luminescence signal by using lab-made optical imaging equipment of animal light imaging system (ALIS) and different two kinds of light sources. One is three bacterial light-emitting sources containing different number of bacteria. The other is three different non-bacterial light sources emitting very weak light. By using the concept of the candela and the flux, we could derive simplified linear quantification formula. After experimentally measuring light intensity, the data was processed with the proposed quantification function. Results: We could obtain linear response of photon counts to measurement time by applying the pre-determined quantification function. The ratio of the re-calculated photon counts and measurement time present a constant value although different light source was applied. Conclusion: The quantification function for linear response could be applicable to the standard quantification process. The proposed method could be used for the exact quantitative analysis in various light imaging equipments with presenting linear response behavior of constant light emitting sources to measurement time.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.2
• /
• pp.123-129
• /
• 2017

This paper proposes a new nonnegative matrix factorization (NMF) based direction-of-arrival (DOA) estimation method for multiple sound sources using a dual microphone array. First of all, sound signals coming from the dual microphone array are segmented into consecutive analysis frames, and a steered-response power phase transform (SRP-PHAT) beamformer is applied to each frame so that stereo signals of each frame are represented in a time-direction domain. The time-direction outputs of SRP-PHAT are stored for a pre-defined number of frames, which is referred to as a time-direction block. Next, In order to estimate DOAs robust to noise, each time-direction block is normalized along the time by using a block subtraction technique. After that, an unsupervised NMF method is applied to the normalized time-direction block in order to cluster the directions of each sound source in a multiple sound source environments. In particular, the activation and basis matrices are used to estimate the number of sound sources and their DOAs, respectively. The DOA estimation performance of the proposed method is evaluated by measuring a mean absolute error (MAE) and the standard deviation of errors between the oracle and estimated DOAs under a three source condition, where the sources are located in [$-35{\circ}$, 5m], [$12{\circ}$, 4m], and [$38{\circ}$, 4.m] from the dual microphone array. It is shown from the experiment that the proposed method could relatively reduce MAE by 56.83%, compared to a conventional SRP-PHAT based DOA estimation method.

• Journal of the Korean earth science society
• /
• v.39 no.5
• /
• pp.501-515
• /
• 2018

In this study, the method of multiple working hypotheses (MMWH) as a method of earth scientific inquiry was applied in a context of abductive reasoning about the formation of a rock with a specific structure, and the characteristics of MMWH revealed in the reasoning process were explored. Participants were 31 senior undergraduate students enrolled in a course in a university of education. As part of the course, the participants performed abductive inquiry with multiple working hypotheses about the formation of a rock. The students were asked to record both the processes and results of their reasoning in sketchbooks. The content of the students' sketchbook reports was analyzed according to the principle of analytic induction. Results demonstrated four assertions. First, the participants' working hypotheses were suggested in the use of resource models, and the adaption of the resource models often occurred in this process. Second, the perceptual properties of evidence influenced the activation of the resource models. Third, the kinds of observed evidence and the different interpretations of evidence resulted into different judgments on working hypotheses. Fourth, sometimes new hypotheses were generated by the combination of alternative hypotheses. Implications of these findings for earth science education and relevant research were discussed.

• Elastomers and Composites
• /
• v.19 no.1
• /
• pp.13-31
• /
• 1984

The purposes of this dissertation are to demonstrate that DSC theromoanaytical methods of vulcanization can provide useful informations on the vulcanization characteristics of industrial formulations and also provides the potential basis for a rapid and complete method of sulfur and vulcanizing accelerator analysis for quality control. The influences of those factors such as heating rate, scan temperature, vucanizing accelerator's type and concentration upon vulcanization exotherm in NR and NBR compounds in the presence of vulcanizing accelerators such as TMTD,MBTS,DPG,TMTM,CBS, and MBT were evaluated by means of DSC. In order to examine the credibility in the DSC method, the same samples which were used for DSC method were studied to compare the DSC results with the ODR (Oscillating Disk Rheometer) data. The results obtained were as follows 1. In the DSC dynamic experiments, the observed enthalpy results from vulcanization depends upon the heating rate; In the range of 2 to $20^{\circ}C/min$ of heating rate, as the heating rate was increased the enthalpy change was also increased. However, over the heating rate of $30^{\circ}C/min$ it was observed that the enthalpy change was decreased as the heating rate was further increased. Without regard to the change of enthalpy, tremendous instantaneous heat evolving was observed in the range of high heating rates. 2. For the samples which are added with various vulcanizing accelerators, the activation energies of vulcanization were as follows; 3. Regarding to the influences of vulcanizining accelerator's types upon the characteristics of vulcanization exotherm, NR and NBR compounds in the presence of thiuramsulphide compounds type accelerators such as TMTD, TMTM, were exhibited sharper and higher vulclanization exotherm than others. From the resuts of DSC thermograms which was distributed in even shape in the broad temperature range, it was clearly shown that the guanidine compounds type accelerator such as DPG acts as a delayed acting accelerator. 4. In the comparison of DSC and ODR results, the dependency of temperature in the cure rate and the observed conversions show good agreements between two results. 5. In the same curatives, by the comparison of glass transition temperatures, it was possible to predict relative values of maximum torques. Consequently, from the present studies, it is shown that the DSC thermoanalytical method can provide an alternate new method for rapid and complete quality control analysis of rubber industry.

• Journal of Food Hygiene and Safety
• /
• v.34 no.3
• /
• pp.269-276
• /
• 2019

A colorimetric rapid detection method based on acetylcholinesterase (AChE) was developed for the analysis of organophosphorus (OP) and carbamate (CB) pesticides. The AChE catalyzes acetylthiocholine into thiocholine having (-) and (+) charges, and the (+) charge results in gold nanoparticle (GNP) aggregation. The in-activation of AChE by OP and CB has been well known. In order to optimize the colorimetric method, optimal dilution times of commercial serum containing AChE, diameter of GNP, and concentration of acetylthiocholine were tested as a key parameter. The colorimetric detection limits of the method were 7.5 ng/mL for both dimethyl amine and carbofuran pesticides in 60% ethanol. No cross-reaction to other chemicals, such as aflatoxin B1 and ochratoxin A, which can be contaminated with pesticides in agricultural products, was observed. Recoveries from lettuce, sesame leaf, and cabbage lettuce spiked with known concentrations of dimethyl amine and carbofuran were found to be ranged from 83.85 to 133.16%. These results indicated that the colorimetric rapid method based on AChE can be a useful tool for the sensitive, specific, rapid, and accurate detection of OP and CB pesticides in fresh vegetables.

• BMB Reports
• /
• v.30 no.5
• /
• pp.303-307
• /
• 1997

Grb2, which is composed of a Src homology 2 (SH2) domain and two Src homology 3 (SH3) domains, is known to serve as an adaptor protein in signaling for Ras activation. Thus, a blocker of the Grb2 interactions with other proteins can be a potential candidate for an anticancer drug. In this study, we have developed a high throughput screening method for SH3 domain binding ligands and blockers. Firstly, we made and purified the glutathione S-transferase (GST)-fusion proteins with the Grb2 SH2 and SH3 domains, and the entire Grb2. This method measures the binding of a biotin-labeled oligopeptide, derived from a Grb2/SH3 binding motif in the hSos, to the GST-fusion proteins, which are precoated as glutathione S-transferase fusion protein on a solid phase. When $1\;{\mu}g$ of each fusion protein was used to coat the wells, both N- and C- terminal SH3 the domains as well as the whole of Grb2 were able to interact with the biotin-conjugated ligand peptide, while the SH2 domain and GST alone showed no binding affinity. Although N- and C- terminal SH3 domains showed an increase of binding to the ligand peptide in proportion to the amount of peptide, the GST fusion protein with Grb2 demonstrated much higher binding affinity. GST-Grb2 coating on the solid phase showed a saturation curve; 66 and 84% of the maximal binding was observed at 100 and 300 ng/$100\;{\mu}l$, respectively. This binding assay system was peptide sequence-specific, showing a dose-dependent inhibition with the unlabeled peptide of SH3 binding motif. Several other peptides, such as SH2 domain binding motifs and PTB domain binding motif, were ineffective to inhibit the binding to the biotin-conjugated ligand peptide. These results suggest that our method may be useful to screen for new anticancer drug candidates which can block the signaling pathways mediated by SH3 domain binding.

• Environmental Mutagens and Carcinogens
• /
• v.26 no.2
• /
• pp.53-58
• /
• 2006

Normally, environmental toxicants are classified as endocrine disruptors if they interfere with regulation of cellular function by endogeneous steroids through inhibition of receptor binding and/or transcriptional activation. So, many studies have been performed about agonist/antagonist of hormone receptor to study mechanisms of endocrine disruptors. If toxicants affect steroid biosynthesis and/or degradation and alter hormone homeostasis, these also are classified as endocrine disruptors. But there are not many studies of the mechanisms of endocrine disruptors on the basis of alteration of steroid biosynthesis and/or degradation. Isolation and culture of Leydig cells from testis is one of methods for the steroidogenesis screening assays to evaluate a substance for altering steroidogenesis. Leydig cells were harvested using the method described by Klinefelter with modifications. Leydig cells were purified by perfusion of testis and incubation ($34^{\circ}C$, 80cycles/minute, 20 minutes) with collagenase (0.25 mg/kg), centrifugal elutriation, percoll gradient centrifugation and BSA multidensity gradient centrifugation. To confirm if this method is one of appropriate tools to evaluate a substance for altering steroidogenesis, ketoconazole, positive control was administered to purified Leydig cells. Ketoconazole ($10^{-8}M$ and above) significantly reduced testosterone production in purified Leydig cells. From above results, we suggest that this method for steroidogenesis screening assay appears to be a appropriate tool to detect suspected compounds for altering steroidogenesis.

• Journal of the Korea Institute of Building Construction
• /
• v.20 no.2
• /
• pp.171-181
• /
• 2020

Recently, free-formed construction technology is becoming a new measure of representing technological superiority and sociocultural ingenuity. However, the CNC processing technology utilizing the existing wood and iron form has limitations in terms of the manufacturing time and material cost. Therefore, in this study, the method and process of manufacturing free-formed EPS form using S-LOM-based 3D printing technology were suggested. Furthermore, through the mock-up test, a comparative analysis of the manufacturing time and precision with CNC milling technology was conducted. The results show that S-LOM-based 3D printing technology has reduced manufacturing time about 57.4% compared to CNC milling technology during the free-formed EPS form manufacturing process. In addition, compared to the design drawings, the maximum error value was 20.5mm, proving the applicability of S-LOM-based 3D printing technology. The results of this study are expected to contribute to the improvement of S-LOM method and the activation of S-LOM method by verifying the applicability of S-LOM-based 3D printing technology.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.8
• /
• pp.1503-1517
• /
• 1994

Intelligent Motion Planning System(IMPS) is presented for a robot to achieve an efficient path toward the given target point in two dimensional unknown environment is constructed with unrestricted obstacle shapes. IMPS consists of three components for making intelligent motion. These components are real-time motion planning algorithm based on a discontinous boundary method, fuzzy neural network decision system for heuristic knowledge representation, and world modeling with forgetting and reinforcing memory cells. First of all, in real-time motion planning algorithm, the behavior-based architectural method is used to generate subgoal. A behavior generates a subgoal independently by using the method of discontinuous boundary in sensed area. The discontinuous boundary method is a new proposed fast obstacle avoidance algorithm. The second component is fuzzy neural network decision system for accomplishing the subgoal. The heuristic rules are imbedded on the fuzzy neural network to make an intelligent decision. The last one is a forgetting, reinforcing memory technique for the construction of external world map. The activation values of all activated memory cells in grid space are decreased monotonically and after all they are burned out. Therefore, after sufficient journey, robot can have a stationary world map even if the dynaic obstacles exist. Using the IMPS, several simulations show the efficient achievement of target point in unknown enviroment with obstcles of various shapes.