Loading [MathJax]/jax/output/CommonHTML/jax.js
  • Title/Summary/Keyword: Classical model

Search Result 1,049, Processing Time 0.024 seconds

Anomaly Detection for User Action with Generative Adversarial Networks (적대적 생성 모델을 활용한 사용자 행위 이상 탐지 방법)

  • Choi, Nam woong;Kim, Wooju
    • Journal of Intelligence and Information Systems
    • /
    • v.25 no.3
    • /
    • pp.43-62
    • /
    • 2019
  • At one time, the anomaly detection sector dominated the method of determining whether there was an abnormality based on the statistics derived from specific data. This methodology was possible because the dimension of the data was simple in the past, so the classical statistical method could work effectively. However, as the characteristics of data have changed complexly in the era of big data, it has become more difficult to accurately analyze and predict the data that occurs throughout the industry in the conventional way. Therefore, SVM and Decision Tree based supervised learning algorithms were used. However, there is peculiarity that supervised learning based model can only accurately predict the test data, when the number of classes is equal to the number of normal classes and most of the data generated in the industry has unbalanced data class. Therefore, the predicted results are not always valid when supervised learning model is applied. In order to overcome these drawbacks, many studies now use the unsupervised learning-based model that is not influenced by class distribution, such as autoencoder or generative adversarial networks. In this paper, we propose a method to detect anomalies using generative adversarial networks. AnoGAN, introduced in the study of Thomas et al (2017), is a classification model that performs abnormal detection of medical images. It was composed of a Convolution Neural Net and was used in the field of detection. On the other hand, sequencing data abnormality detection using generative adversarial network is a lack of research papers compared to image data. Of course, in Li et al (2018), a study by Li et al (LSTM), a type of recurrent neural network, has proposed a model to classify the abnormities of numerical sequence data, but it has not been used for categorical sequence data, as well as feature matching method applied by salans et al.(2016). So it suggests that there are a number of studies to be tried on in the ideal classification of sequence data through a generative adversarial Network. In order to learn the sequence data, the structure of the generative adversarial networks is composed of LSTM, and the 2 stacked-LSTM of the generator is composed of 32-dim hidden unit layers and 64-dim hidden unit layers. The LSTM of the discriminator consists of 64-dim hidden unit layer were used. In the process of deriving abnormal scores from existing paper of Anomaly Detection for Sequence data, entropy values of probability of actual data are used in the process of deriving abnormal scores. but in this paper, as mentioned earlier, abnormal scores have been derived by using feature matching techniques. In addition, the process of optimizing latent variables was designed with LSTM to improve model performance. The modified form of generative adversarial model was more accurate in all experiments than the autoencoder in terms of precision and was approximately 7% higher in accuracy. In terms of Robustness, Generative adversarial networks also performed better than autoencoder. Because generative adversarial networks can learn data distribution from real categorical sequence data, Unaffected by a single normal data. But autoencoder is not. Result of Robustness test showed that he accuracy of the autocoder was 92%, the accuracy of the hostile neural network was 96%, and in terms of sensitivity, the autocoder was 40% and the hostile neural network was 51%. In this paper, experiments have also been conducted to show how much performance changes due to differences in the optimization structure of potential variables. As a result, the level of 1% was improved in terms of sensitivity. These results suggest that it presented a new perspective on optimizing latent variable that were relatively insignificant.

Condensation processes in transonic two-phase flows of saturated humid air using a small-disturbance model (미교란 모델을 이용한 포화 습공기 천음속 2상 유동에서의 응축현상)

  • Lee, Jang-Chang;Zvi Rusak
    • Journal of the Korean Society for Aeronautical & Space Sciences
    • /
    • v.31 no.6
    • /
    • pp.23-29
    • /
    • 2003
  • Transonic two-phase flow of Saturated humid air, in which relative humidity is 100%, with various condensation processes around thin airfoils is investigated. The study uses an extended transonic small-disturbance(TSD) model of Rusak and Lee [11, 12] which includes effects of heat addition to the flow due to condensation. Two possible limit types of condensation processes are considered. In the nonequilibrium and homogeneous process, the condensate mass fraction is calculated according to classical nucleation and droplet growth rate models. In the equilibrium process, the condensate mass fraction is calculated by assuming an isentropic process. The flow and condensation equations are solved numerical1y by iterative computations. Results under same upstream conditions describe the flow structure, field of condensate, and pressure distribution on airfoil's surfaces. It is found that flow characteristics, such as position and strength of shock waves and airfoil’s pressure distribution, are different for the two condensation processes. Yet, in each case, heat addition as a result of condensation causes significant changes in flow behavior and affects the aerodynamic performance of airfoils.

Gene Co-expression Network Analysis Associated with Acupuncture Treatment of Rheumatoid Arthritis: An Animal Model

  • Ravn, Dea Louise;Mohammadnejad, Afsaneh;Sabaredzovic, Kemal;Li, Weilong;Lund, Jesper;Li, Shuxia;Svendsen, Anders Jorgen;Schwammle, Veit;Tan, Qihua
    • Journal of Acupuncture Research
    • /
    • v.37 no.2
    • /
    • pp.128-135
    • /
    • 2020
  • Background: Classical acupuncture is being used in the treatment of rheumatoid arthritis (RA). To explore the biological response to acupuncture, a network-based analysis was performed on gene expression data collected from an animal model of RA treated with acupuncture. Methods: Gene expression data were obtained from published microarray studies on blood samples from rats with collagen induced arthritis (CIA) and non-CIA rats, both treated with manual acupuncture. The weighted gene co-expression network analysis was performed to identify gene clusters expressed in association with acupuncture treatment time and RA status. Gene ontology and pathway analyses were applied for functional annotation and network visualization. Results: A cluster of 347 genes were identified that differentially downregulated expression in association with acupuncture treatment over time; specifically in rats with CIA with module-RA correlation at 1 hour after acupuncture (-0.27; p < 0.001) and at 34 days after acupuncture (-0.33; p < 0.001). Functional annotation showed highly significant enrichment of porphyrin-containing compound biosynthetic processes (p < 0.001). The network-based analysis also identified a module of 140 genes differentially expressed between CIA and non-CIA in rats (p < 0.001). This cluster of genes was enriched for antigen processing and presentation of exogenous peptide antigen (p < 0.001). Other functional gene clusters previously reported in earlier studies were also observed. Conclusion: The identified gene expression networks and their hub-genes could help with the understanding of mechanisms involved in the pathogenesis of RA, as well understanding the effects of acupuncture treatment of RA.

Study on Vehicle Routing Problem with Minimum Delivery Completion Time (특송소화물 배송완료시간 최소화를 위한 차량경로문제 연구)

  • Lee, Sang-Heon
    • Journal of Korean Society of Transportation
    • /
    • v.22 no.7 s.78
    • /
    • pp.107-117
    • /
    • 2004
  • The growing demand for customer-response, made-to-order manufacturing and satisfactory delivery are stimulating the importance of commercial fleet management problem. Moreover, the rapid transformation to the customer-oriented multi-frequency, relatively small fleet, such as home delivery and Perishable goods, requiring prompt delivery and advanced real-time operation of vehicle fleets. In this paper we consider the vehicle routing problem(VRP) to minimize delivery completion time which is equal to the time that last customer wait for the vehicle in fleet operation. The mathematical formulation is different from those for the classical VRP which is minimizing cost/distance/time by running vehicles in manager's point of view. The key aspect of this model is not considering the return time from the last customer to depot in every vehicle path. Thereby, the vehicle dispatcher can afford to dynamically respond to customer demand and vehicle availability. The customer's position concerned with minimizing waiting time that may be applied for the delivery of product required freshness or delivery time. Extensive experiments are carried out to compare the performance of minimizing delivery completion time by using the ILOG Solver which has the advantage of solving quickly an interim solution very near an optimal solution. The experimental results show that the suggested model can easily find near optimal solution in a reasonable computational time under the various combination of customers and vehicles.

A Development of Pre and Post Processor for Design of Surface System of Free Form Building (비정형 건축물의 외피시스템 설계를 위한 전·후처리 모듈 개발)

  • Park, Se-Hee;Jung, Sung-Jin;Lee, Jae-Sung
    • Journal of the Korea Academia-Industrial cooperation Society
    • /
    • v.19 no.9
    • /
    • pp.333-340
    • /
    • 2018
  • Recently, free-form buildings have been designed with complex shapes due to digitization of the construction industry. Exterior and interior components of free-form buildings have free cross sections and curved shapes. Therefore, structural members with curvature are frequently seen. In the modeling and stability evaluation of these structures, commercial programs using classical finite element analysis are not able to perform rapid shape modeling, resulting in a decrease in productivity. Therefore, in this study, pre- and post-processing modules were developed using a prior study to rapidly model the surface of a free-form building and to automatically generate frame structures that make up the cladding. The developed modules use a subdivision algorithm with spline curves. This algorithm is used to automatically generate analytical elements from the configuration information of NURBS curves. In addition, the deformation after analysis can be viewed more realistically. The modules can quickly construct complex curved surfaces. An analysis model of the frame structure was also automatically generated. Therefore, the modules could contribute to the productivity improvement of free-form building design.

Nerve Regeneration Using a Vein Graft Conduit filled with Hyaluronic Acid in a Rat Model (흰쥐 모델에서 하이알루론산을 채운 정맥도관의 신경재생에 관한 연구)

  • Suh, Bo Ik;Kim, Sang Woo;Chung, Ho Yun;Kim, Il Hwan;Yang, Jung Dug;Park, Jae Woo;Cho, Byoung Chae
    • Archives of Plastic Surgery
    • /
    • v.34 no.3
    • /
    • pp.279-284
    • /
    • 2007
  • Purpose: The vein graft was considered as a useful conduit for nerve defect. But the problem is that it might be collapsed in long vein graft state. A new experimental model using vein graft filled with hyaluronic acid was considered. Methods: Thirty rats were used for the experimental animal. In group I, one side of the femoral nerve was exposed and a segment was removed about 15mm. The neural gap was connected with nerve graft. In group II, the nerve gap was connected with vein graft only. In group III, the nerve gap was connected with vein graft filled with hyaluronic acid. A walking track analysis was made periodically for 2 months and NCV(nerve conduction velocity) was executed at the end of the experiment. And morphologic studies were also done for all groups Results: In a walking track analysis, the toe-spread was widen and the foot-length was lengthened. The recovery of the toe-spread and foot length was checked 2 weeks interval, periodically for two months. The SFI (sciatic function index) was 52.5±8.2 in group I, 68.1±4 in group II, 55.3±7.9 in group III. In electrophysiological study, NCV(nerve conduction velocity) was 26.71±3.11m/s in group I, 17.94±4.35m/s in group II, 25.69±2.81m/s in group III. The functional recovery in group I and III was superior to that the group II statistically(p < 0.05) Under electromicroscopic study, the number of the myelinated axons were 1419.1±240 in group I, 921.7±176.8 in group II, 1322.2±318 in group III. The number of the myelinated axons were much more in group I and III than group II statistically (p<0.05). Conclusion: This study suggested that the vein graft filled with hyaluronic acid is more effective than vein graft only for the conduit of the nerve gap. It was thought that the technique could be used in clinical cases with nerve defects as an alternative method to classical nerve grafts.

A Route Search of Urban Traffic Network using Fuzzy Non-Additive Control (퍼지 비가법 제어를 이용한 도시 교통망의 경로 탐색)

  • 이상훈;김성환
    • Journal of Korean Society of Transportation
    • /
    • v.21 no.1
    • /
    • pp.103-113
    • /
    • 2003
  • This paper shows alternative route search and preference route search for the traffic route search, and proposes the use of the fuzzy non-additive controller by the application of AHP(analytic hierarchy process). It is different from classical route search and notices thinking method of human. Appraisal element, weight of route is extracted from basic of the opinion gathering for the driving expert and example of route model was used for the finding of practice utility. Model evaluation was performed attribute membership function making of estimate element, estimate value setting, weight define by the AHP, non additive presentation of weight according to λ-fuzzy measure and Choquet fuzzy integral. Finally, alternative route search was possible to real time traffic route search for the well variable traffic environment, and preference route search showed reflection of traffic route search disposition for the driver individual. This paper has five important meaning. (1)The approach is similar to the driver's route selection decision process. (2)The approach is able to control of route appraisal criteria for the multiple attribute. (3)The approach makes subjective judgement objective by a non additive. (4)The approach shows dynamic route search for the alternative route search. (5)The approach is able to consider characteristics of individual drivers attributed for the preference route search.

Sintering Mixtures in the Stage of Establishing Chemical Equilibrium

  • Savitskii, A.P.
    • Proceedings of the Korean Powder Metallurgy Institute Conference
    • /
    • 1999.04a
    • /
    • pp.5-5
    • /
    • 1999
  • The Principal deficiency of the existing notion about the sintering-mixtures consists in the fact that almost no attention is focused on the Phenomenon of alloy formation during sintering, its connection with dimensional changes of powder bodies, and no correct ideas on the driving force for the sintering process in the stage of establishing chemical equilibrium in a system are available as well. Another disadvantage of the classical sintering theory is an erroneous conception on the dissolution mechanism of solid in liquid. The two-particle model widely used in the literature to describe the sintering phenomenon in solid state disregards the nature of the neighbouring surrounding particles, the presence of pores between them, and the rise of so called arch effect. In this presentation, new basic scientific principles of the driving forces for the sintering process of a two-component powder body, of a diffusion mechanism of the interaction between solid and liquid phases, of stresses and deformation arising in the diffusion zone have been developed. The major driving force for sintering the mixture from components capable of forming solid solutions and intermetallic compounds is attributed to the alloy formation rather than the reduction of the free surface area until the chemical equilibrium is achieved in a system. The lecture considers a multiparticle model of the mixed powder-body and the nature of its volume changes during solid-state and liquid-phase sintering. It explains the discovered S-and V-type concentration dependencies of the change in the compact volume during solid-state sintering. It is supposed in the literature that the dissolution of solid in liquid is realised due to the removal of atoms from the surface of the solid phase into the melt and then their diffusicn transfer from the solid-liquid interface into the bulk of liquid. It has been shown in our experimental studies that the mechanism of the interaction between two components, one of them being liquid, consist in diffusion of the solvent atoms from the liquid into the solid phase until the concentration of solid solutions or an intermetallic compound in the surface layer enables them to pass into the liquid by means of melting. The lecture discusses peculimities of liquid phase formation in systems with intermediate compounds and the role of the liquid phase in bringing about the exothermic effect. At the frist stage of liquid phase sintering the diffusion of atoms from the melt into the solid causes the powder body to grow. At the second stage the diminution of particles in size as a result of their dissolution in the liquid draws their centres closer to each other and makes the compact to shrink Analytical equations were derived to describe quantitatively the porosity and volume changes of compacts as a result of alloy formation during liquid phase sinteIing. Selection criteria for an additive, its concentration and the temperature regime of sintering to control the density the structure of sintered alloys are given.

  • PDF

A Numerical Method for Nonlinear Wave-Making Phenomena (비선형 조파현상의 수치해법)

  • Jang-Whan Kim;Kwang-June Bai
    • Journal of the Society of Naval Architects of Korea
    • /
    • v.30 no.1
    • /
    • pp.65-72
    • /
    • 1993
  • A numerical method for nonlinear free-surface-wave problem is developed in this paper. The final goal of this study is to simulate the towing tank experiment of a ship model and to partially replace the experiment by the numerical model. The exact problem in the scope of potential flow theory is formulated by a variational principle based on the classical Hamilton's principle. A localized finite element method is used in the present numerical computations which made use of the following two notable steps. The first step is an efficient treatment of the numerical radiation condition by using the intermediate nonlinear-to-linear transition buffer subdomain between the fully nonlinear and linear subdomains. The second is the use of a modal analysis in the final stage of the solution procedures, which enables us to reduce the computation time drastically. With these improvements the present method can treat a much larger computational domain than that was possible previously. A pressure patch on the free surface was chosen as an example. From the present computed results we could investigate the effect of nonlinearity on the down-stream wave pattern more clearly than others, because much larger computational domain was treated. We found, specifically, the widening of the Kelvin angle and the increase of the wave numbers and the magnitude of wave profiles.

  • PDF

Endless debates on the extant basal-most angiosperm (현생 기저 피자식물에 대한 끝나지 않는 논쟁)

  • Kim, Sangtae
    • Korean Journal of Plant Taxonomy
    • /
    • v.40 no.1
    • /
    • pp.1-15
    • /
    • 2010
  • Recognizing a basal group in a taxon is one of the most important factors involved in understanding the evolutionary history of that group of life. Many botanists have suggested a sister to all other angiosperms to understand the origin and rapid diversification of angiosperms based on morphological and fossil evidence. Recent technical advances in molecular biology and the accumulation of molecular phylogenetic data have provided evidence of the extant basal-most angiosperm which is a sister to all other angiosperms. Although it is still arguable, most plant taxonomists agree that Amborella trichopoda Baill., a species (monotypic genus and monotypic family) distributed in New Caledonia, is a sister to all other extant angiosperms based on evidence from the following molecular approaches: 1) classical phylogenetic analyses based on multiple genes (or DNA regions), 2) analyses of a tree network of duplicated gene families, and 3) gene-structural evidence. As an alternative hypothesis with relatively minor evidence, some researchers have also suggested that Amborella and Nymphaeaceae form a clade that is a sister to all other angiosperms. Debate regarding the basal-most angiosperms is still ongoing and is currently one of the hot issues in plant evolutionary biology. We expect that sequencing of the whole genome of Amborella as an evolutionary model plant and subsequent studies based on this genome sequence will provide information regarding the origin and rapid diversification of angiosperms, which is Darwin's so called abominable mystery.